Post on 22-Aug-2020
UNIVERSITÀ DEGLI STUDI DI MILANO
Facoltà di Scienze MM.FF.NN.
Dipartimento di Chimica Fisica ed Elettrochimica
Corso di Dottorato di Ricerca in SCIENZE CHIMICHE - CHIM/02
CHALLENGING QUESTIONS IN CHEMISTRY: THE SYNERGIC ROLE OF AB-INITIO CALCULATIONS, X-RAY DERIVED CHARGE DENSITIES AND MODERN
TOPOLOGICAL APPROACHES TO BONDING
Relatore: Dr. Leonardo Lo Presti Correlatore: Dr. Carlo Gatti Coordinatore del Corso di Dottorato: Prof. Silvia Ardizzone Tesi di Dottorato di Ricerca di: Simone Cenedese Matr. R08325
Anno Accademico 2010-2011
Index
Chapter 1: Introduction pag. 1
Chapter 2: FeX2 (X = S, P, As, Sb) marcasitic compounds pag. 3
1. Introduction pag. 3
2. Marcasites crystallographic structure pag. 3
2.1. Local geometric structure, d-orbital splitting and formal charge of iron pag. 6
3. Computational details pag. 9
4. Results pag. 11
4.1. Electronic structure: dependency from the level of theory pag. 11
4.2. Electronic structure: orbital-based analysis pag. 18
4.3. Electronic structure: observable-based analysis pag. 22
5. Conclusions pag. 35
Chapter 3: Hypervalency in K2SO4 pag. 38
1. Introduction pag. 38
2. Computational details pag. 41
3. Results and Discussion pag. 41
4. Conclusions pag. 54
Chapter 4: Coordination Polymers: the case of [Zn(HCOO)2(H2O)2]∞∞∞∞ pag. 56
1. Introduction pag. 56
2. Crystallographic structure pag. 57
3. Computational details pag. 58
4. Results and Discussion pag. 59
5. Conclusions pag. 65
Chapter 5: Conclusions and Future Work pag. 67
References pag. 69
1
Chapter 1
Introduction
During the last 20 years the analysis of the charge density has grown from a small area of research
reserved for scientists almost exclusively devoted to this field to quite a routine technique. Many
books have been devoted to this field and it is common nowadays working with instruments and
program packages designed for collect, refine and model the electron density almost without the
help of an external user. In some cases, the same can be said also when thinking at the charge
density derived from ab-initio calculations.
Nonetheless, many issues still exist that makes this analysis less of a routine a more real scientific
work. Thinking at the problems one can encounter when performing XRD measurements and
refining diffraction data, it is possible to cite the presence of heavy elements in the structure, high
symmetry, extinction effects, static and dynamic disorder. There are cases in which these problems
can be overcome or, at least, minimized, but still there’s a huge number of chemical systems for
which ambiguities and (sometimes) contradictory outcomes result from the experimental analysis
alone (especially for what concerns bonding and not just structure).
This work deals with the role ab-initio calculations have in supporting and, when needed, helping to
clarify the meaning of experimental findings. Furthermore, the extra information one gets from the
wavefunction, that is the knowledge of the density matrix and of the pair density, both customarily
unavailable from experiment, has revealed of uttermost importance in the study of the challenging
chemical bond features investigated in this thesis.
This work is the result of a strong and still ongoing collaboration with the “Charge Density” group
at the University of Århus (Denmark), whose members were responsible of the collection,
reduction, and analysis of the experimental data. In particular, PhD student Mette S. Schmøkel
2
followed the work on K2SO4 and is currently working on FeX2 marcasitic systems, and PhD student
Mads R. V. Jørgensen the work on transition metal formates.
3
Chapter 2
FeX2 (X = S, P, As, Sb) marcasitic compounds
1. Introduction
Marcasite type structures have recently received huge interest since it was discovered that FeSb2
exhibits a colossal thermoelectric power factor at low temperature,1,2 and if the thermal conductivity
can be lowered the material holds the potential for providing an efficient cooling technology e.g. for
superconducting transmission cables. The extraordinary physical properties are still not well
understood3 and clearly it is of fundamental interest to examine in depth the crystal structure,
electronic structure and chemical bonding.
In the next paragraphs, the electronic structure for marcasites (M) and löllingites (L) is revised and
it will be shown how geometry, electronic structure and bonding interactions are intriguingly
interplayed in such compounds. The predictions of ligand field theory (LFT) are compared with the
ab-initio orbital-based view of both the anisotropy of the iron d-electrons distribution and the
composition of bonding and antibonding states close to the Fermi level. The orbital-based picture is
tested against the study of the charge transfer, the nature of the bonding interactions and the valence
anisotropy of the iron atom using the topology of two observables, the electron density, ρ and its
Laplacian, ∇2 ρ within the Quantum Theory of Atoms In Molecules (QTAIM).
2. Marcasites crystallographic structure
The crystallographic structures of marcasite type compounds were extensively scrutinized in a
series of studies around 1970.4, 5, 6, 7, 8 The binary compounds TX2 (T = transition metal, X group
XV or XVI element) that belong to the marcasite type structure crystallize in the Pnnm space group
(#58) and have two formula units in the unit cell (Figure 2.1). Within this crystal structure,
4
however, two distinct groups of compounds, sharing the same space group, but distinguished
according to the ratios c/a and c/b of the cell parameters, can be identified. Marcasites (M), for
which the reference compound is FeS2, exhibit c/a and c/b values close to 0.74 and 0.62,
respectively. On the other hand, significantly lower ratios, typically 0.55 (c/a) and 0.48 (c/b), are
found in löllingites (L), whose standard reference compound is FeAs2. Other definitions, reported in
Table 2.1, appeared in literature from time to time. The terms marcasite and löllingite refer to the
mineralogical definition and will be adopted in the following.
Figure 2.1 TX2 marcasite type crystal structure. Left: each T atom (blue) is coordinated with six X atoms (yellow) in an octahedral environment; each X atom is surrounded by three T and one X atoms placed at the corners of a tetrahedron (shown only for the X atom in the top-left part of the cell, for the sake of clarity). Right: TX6 octahedra are connected along a and b axes through X2 dumbbells (only those along b are shown in the figure), while along c they are edge-sharing.
Table 2.1 Proposed nomenclature, based on the basis of the cell parameters (c/a and c/b) ratio, for the two different groups of compounds belonging to the marcasite type structure.
c/a c/b Definitions
0.74 0.62 marcasites
anomalous marcasites normal marcasites
0.55 0.48
löllingites Jahn-Teller marcasites compressed marcasites
regular marcasites
5
Two main differences can thus be highlighted between L and M types of compounds: a geometric
one, due to the notable compression along the c axis in the former, and an electronic one, given by
the two less valence electrons per formula unit characterizing L rather than M type structure.
Iron disulphide is the leading term of marcasites and the M structure is retained also when sulphur is
replaced with other group VI elements like Se and Te. When substituting chalcogen atoms with
group V ones, i.e. removing 2 e-/formula unit, the löllingite structure comes into being (Table 2.2).
Table 2.2 TX2 (T = Fe, Ni ; X = P, As, Sb, S, Se, Te) marcasite- and löllingite-type compounds. Cell parameters a, b and c are given in Ångstrom, while the Fe-X-Fe angle along the c axis (δ, see also Figure 2.3) is given in degrees. The cell parameters (c/a and c/b) ratios are also shown for each compound in the table.
Marcasite-type Löllingite-type
FeS26,8
a = 4.4431 Å b = 5.4245 Å c = 3.3871 Å
c/a = 0.76 c/b = 0.62 δ = 97.4°
FeP25
a = 4.9729 Å b = 5.6568 Å c = 2.7230 Å
c/a = 0.55 c/b = 0.48 δ = 73.9°
FeSe29
a = 4.8002 Å b = 5.7823 Å c = 3.5834 Å
c/a = 0.75 c/b = 0.62 δ = 97.5°
FeAs29
a = 5.3012 Å b = 5.9858 Å c = 2.8822 Å
c/a = 0.54 c/b = 0.48 δ = 74.3°
FeTe210,6
a = 5.2655 Å b = 6.2679 Å c = 3.8738 Å
c/a = 0.74 c/b = 0.62 δ = 97.8°
FeSb210,11
a = 5.8328 Å b = 6.5376 Å c = 3.1973 Å
c/a = 0.55 c/b = 0.49 δ = 76.0°
NiAs29
a = 4.7582 Å b = 5.7949 Å c = 3.5440 Å
c/a = 0.74 c/b = 0.61 δ = 95.5°
FeAs29
a = 5.3012 Å b = 5.9858 Å c = 2.8822 Å
c/a = 0.54 c/b = 0.48 δ = 74.3°
NiSb29
a = 5.1823 Å b = 6.3168 Å c = 3.8403 Å
c/a = 0.74 c/b = 0.61 δ = 96.7°
FeSb210,11
a = 5.8328 Å b = 6.5376 Å c = 3.1973 Å
c/a = 0.55 c/b = 0.49 δ = 76.0°
This consideration holds also upon substitution of the cation. When Fe is replaced by Ni in FeAs2
and FeSb2 the corresponding marcasites are obtained (Table 2.2). NiP2, on the contrary, crystallize
in the Pa3 space group, retaining the NiS2 pyrite crystal structure. It must be noted however that
pyrite is also the most stable polymorph of FeS2. The removal of just 1 e-/formula unit from the
marcasite structure involves major structural modifications leading to the monoclinic arsenopyrite
structure, space group P21/c (#14). The leading term is the ternary compound FeAsS, clearly
6
highlighting the intermediate composition between FeS2 and FeAs2. However, also the binary ones
exist, having CoSb2 as leading term.
The present study focuses on examining some of the representatives of the marcasites and
löllingites families: FeS2, FeP2, FeAs2 and FeSb2.
2.1. Local geometric structure, d-orbital splitting and formal charge of iron
According to Goodenough, 12 who critically reviewed the conjecture proposed by Hullinger &
Mooser13 and the expansion model proposed by Brostigen & Kjekshus,6 the the electronic structure
of these compounds can be rationalized considering the distorted octahedral environment of the
transition metal. The analysis has been conducted considering mainly FeS2, FeAsS and FeAs2, and
subsequently generalized to TX2 marcasites. It requires a proper choice of the reference system: of
the six Fe-X interatomic distances, two are shorter than the other four: the z axis is aligned along the
short Fe-X interactions (axial) while the x and y axes, which point approximately along the longest
bonds (equatorial), are rotated so as to be equidistant from the internuclear axes of such bonds.
Also in the following, when Mulliken charges and orbital-projected DOS referring to specific
orbitals (like dxy or pz, for example) will be discussed, they were all obtained using the
aforementioned “chemical reference system” for the various compounds.
According to the ligand field theory (LFT), in a perfect octahedral environment the five 3d orbital
of the metal atom are split into the three t2g orbitals (dxy , dyz, dxz) of lower energy and two eg orbitals
(dx2-y2, dz2) of higher energy. This splitting stems from the fact that the latter orbitals point directly
towards the negatively charged ligands whereas the former are directed in between them. In our
case, a further increase in the energy of the dz2 orbital is expected since the axial interactions are
shorter than the equatorial ones. Furthermore, in the equatorial plane, the X ligands approach the Fe
atom with angles different from 90°, so that a lowering of dx2-y2 and a rising of dxy, with respect to
7
the LFT ideal octahedral splitting is expected. These considerations lead to the scheme depicted in
Figure 2.2 for Fe d-orbitals splitting.
As previously mentioned, geometrical distortions lead to x and y axes that are not perfectly aligned
with the Fe-X equatorial interactions (Figure 2.3 and Table 2.3).
Figure 2.2 Splitting scheme for iron d-orbitals in the case of the free atom (left), for a perfect octahedral environment (middle) and in the case of the distorted octahedral field in marcasite and löllingite type of structure (right). The separation between the orbitals is purely qualitative and doesn’t rely on energy calculations.
Figure 2.3 TX6 octahedra structure: left) seen along the c axis: the two axial (ax.) T-X interactions and the tilting (angle ω) of the equatorial plane with respect to them. right) down the c axis: the four equatorial (eq.) T-X interactions and the angle δ, which is related to the compression of the c axis on passing from M to L. See Table 2.3 for numerical values.
8
Table 2.3 FeX2 (X=S,P) lattice and geometrical parameters for the experimental (exp) and optimized (opt) structures valuated with the PBE and B3PW functionals. The values in parentheses (for PBE) refer to the most contracted basis set, optimized for B3PW.
FeS2 FeP2 FeAs2 FeSb2
exp
a (Å) 4.4431 4.9729 5.3012 5.8328 b (Å) 5.4245 5.6568 5.9858 6.5376 c (Å) 3.3871 2.7230 2.8822 3.1973
Fe-X ax. (Å) 2.2359 2.2484 2.3620 2.5746 Fe-X eq. (Å) 2.2539 2.2637 2.3874 2.5965
X-X (Å) 2.2154 2.2365 2.4915 2.8941 δ (degrees) 97.4 73.9 74.3 76.0 ω (degrees) 1.9 1.9 1.9 1.7
opt (PBE)
a (Å) 4.4426
(4.4759) 4.9758
(4.9988) 5.3122
(5.2992) 5.9236
(5.7997)
b (Å) 5.4114
(5.3939) 5.6529
(5.6571) 5.9772
(5.9872) 6.6050
(6.4877)
c (Å) 3.3956
(3.3813) 2.7233
(2.7087) 2.8802
(2.8826) 3.1156
(3.2398)
Fe-X ax. (Å) 2.2311
(2.2240) 2.2444
(2.2460) 2.3625
(2.3584) 2.6216
(2.5654)
Fe-X eq. (Å) 2.2464
(2.2414) 2.2594
(2.2616) 2.3813
(2.3837) 2.5915
(2.5852)
X-X (Å) 2.2613
(2.3036) 2.2610
(2.2681) 2.5135
(2.5148) 2.8930
(2.9065)
δ (degrees) 98.2
(97.9) 74.1
(73.6) 74.4
(74.4) 73.9
(77.6)
ω (degrees) 1.9
(1.7) 1.9
(1.8) 1.7
(1.9) 1.2
(1.4)
opt (B3PW)
a (Å) 4.4392 4.9590 5.2506 5.7538 b (Å) 5.4081 5.6961 6.0333 6.5447 c (Å) 3.4003 2.7117 2.8870 3.2374
Fe-X ax. (Å) 2.2419 2.2747 2.3924 2.6009 Fe-X eq. (Å) 2.2645 2.2461 2.3673 2.5696
X-X (Å) 2.1484 2.2463 2.4801 2.8723 δ (degrees) 97.3 74.3 75.1 78.1 ω (degrees) 1.6 1.9 2.2 1.7
Central to almost any assumption on the electronic structure of these compounds, is that both M and
L structures are characterized by noninteracting diatomic anions (X2)n- . Assuming the presence of
one X-X bond per formula unit, the formal charges of the anion are n = 2 or 4 for chalcogen and
pnicogen ligand, respectively. For iron compounds, this would result in a formal charge of +2 and
9
+4, respectively, for Fe in a fully ionic description. However, this picture is hardly supported by the
known similarity of Pauling electronegativity values (Table 2.4) nor is by the observation that
chalcogen atoms have an higher electronegativity value than pnicogen ones, which makes the
predicted trend of formal charges of iron by the fully ionic description even more unrealistic.
Table 2.4 Pauling electronegativities (χ) and electronegativity differences for Fe, pnicogen (Xpnic. = P, As, Sb) and chalcogen (Xchalc. = S, Se, Te) elements.
χ(Fe) χ(Xpnic.) χ(Fe) - χ(Xpnic.) χ(Xchalc.) χ(Fe) - χ(Xchalc.)
Fe 1.83 P 2.19 0.36 S 2.58 0.75 As 2.18 0.35 Se 2.55 0.72 Sb 2.05 0.22 Te 2.10 0.27
In a recent paper14 a different view was suggested for FeSb2. Instead of considering the Sb2 pairs as
noninteracting units, they could be considered as the rungs of an infinite ladder running trough c
axis in the crystal, the ladder being the result of the coupling of two infinite linear Sb chains due to
Peierls distortion. Doing so results in Sb holding a -1 formal charge and, as a consequence, a more
conventional +2 state for iron, consistent with the 18 electrons rule. The same approach has been
recently adopted for the analysis of MN2 (M = Ir, Os, Rh) transition metal nitrides.15 However, the
authors themselves acknowledge that this analysis cannot be generalized and used to explain the
formal oxidation states of the other compounds. For NiSb2, for example, taking into account Peierls
distortion results again in a +2 formal charge for the metal but then Sb2 pairs would be characterized by a Sb-
Sb double bond. The presence of such a bond, on the other side, in the case of isolated Sb2 pairs would lead
to an unusual +4 oxidation state for Ni.
3. Computational details
The results reported here have been obtained via periodic ab-initio Density Functional Theory
(DFT) calculations employing the CRYSTAL06 code.16 Mainly all electron def2-TZVP basis sets
10
17,18 have been used.a For FeSb2 a description based on small-core (16 and 23 electrons explicitly
described for Fe and Sb respectively) pseudo-potentials has also been considered.19 Due to the
diffuseness of the outer functions of such basis sets, it was not possible to use the original basis set
formulation in the CRYSTAL code. In general, the most diffuse functions were cut while the
Gaussian exponents of the remaining valence functions on iron were recalculated according to the
empirical “1/3” rule [ref. 16, Chapter 7]. Finally, the valence functions have been optimized in the
crystal [ref. 16, Chapter 7] for the various basis sets and the considered DFT functionals, PBE and
B3PW.20,21 This resulted in more contracted basis set to be used with the B3PW functional.
As recommended for heavy atoms,16 extra-large grids (keyword XLGRID in the CRYSTAL code)
were used for the numerical evaluation of the exchange-correlation potential and energy. They
consist of atom-centered contributions with 75 radial points, covering a range of 4.0 ionic radii, and
974 angular points. Wavefunctions, density of states (DOS) and all the other one-electronic
properties obtained from the wavefunction were obtained using a shrinking factor of 12 for
reciprocal space sampling according to the Pack-Monkhorst method, and of 24 for defining the
Gilat net, used in the interpolation of the density matrix and the evaluation of the Fermi energy.
This results in a total of 343 and 2197 sampling points in the Pack-Monkhorst and in the Gilat nets,
respectively. Truncation thresholds were set to 10-9, 10-9, 10-9, 10-9 and 10-18 a.u. (hartree) for the
Coulomb and exchange series and to 10-7 a.u. for the Self Consistent Field (SCF) convergence (total
energy and eigenvalues).
Both the experimental and optimized geometries were considered for each investigated structure. A
full optimization of cell parameters and atomic coordinates was performed within the Pnnm space
group while freezing the cell volume at the experimental values (Table 2.1).
a It should be noted that Sb def2-TZVP basis set is not an “all-electron” since it relies on a pseudo-
potential. For this reason an older TZVP formulation has been considered for Sb basis set.
11
4. Results
The DOS displayed in Figure 2.4 shows that, as expected, the region near the Fermi Level (FL) is
dominated by the interactions between Fe and X, with the contribution from the metal atom being
larger than that from the X atoms for both compounds below the FL. Above, but close to the FL, the
DOS of the conduction bands show an almost similar trend for löllingites, whereas comparable
contributions from Fe and S are found for FeS2. DOS projected on the atomic orbital subsets of Fe
and X (Figure 2.5) reveal that the Fe-X interaction is essentially due to Fe d and X p states for both
FeS2 and FeP2, with the weight of X p states around the FL being smaller for valence than
conduction bands in the former and the opposite being true in the latter compound.
4.1. Electronic structure: dependency from the level of theory
All the studied FeX2 compounds are reported to be small gap diamagnetic conductors. Considering
the estimated band gap values (Table 2.5), when the PBE functional is employed a good agreement
with the reported experimental value (0.37eV) is found for FeP2.22 To our knowledge, no
experimental data are available for FeS2,23 yet our computed value favorably compares with the
previous ab-initio estimate (0.91 eV).2323 For FeAs2 a band gap of 0.20eV is generally reported,24
while for FeSb2 values ranging from 0.3eV to 0.1eV can be found 25, 26, 27, the situation being further
complicated by the presence of smaller indirect gaps, causing a non-zero DOS at the Fermi level.28
This could also be the case, as can be seen from the Density of States (DOS) in Figure 2.4 and,
more clearly, from the bandwidth in Figure 2.6.
12
Figure 2.4 Density Of States (DOS) for the FeX2 (X = S, P, As, Sb) structures at the experimental geometry evaluated with the PBE (left) and B3PW (right) functionals. For each structure, the total and the atom-projected DOS (pDOS) are shown in the top panels. The Fermi Level is indicated by the dashed vertical red line in all plots.
13
Figure 2.5 Projected Density Of States (pDOS) for the FeX2 (X = S, P, As, Sb; from left to right) structures at the experimental geometry evaluated with the PBE functional. For each structure, the top and bottom panels display the DOS projected on the atomic orbital subsets for Fe and X respectively.
14
Figure 2.6 Bandwidth for the FeX2 (X = S, P, As, Sb) structures at the experimental geometry evaluated with the PBE (left) and B3PW (right) functionals.
15
Table 2.5 Band gap values, reported in eV, evaluated for the experimental (exp) and optimized (opt) structures. The two values reported for the PBE basis set refer to the adoption of the two TZ basis sets considered with this functional: the PBE-optimized one (first line, roman) and the B3PW-optimized one (second line, italics)
FeS2 FeP2 FeAs2 FeSb2 exp Opt exp opt exp opt exp opt
PBE 0.83 0.87 0.39 0.41 0.30 0.28 -0.07 -0.07 0.83 0.83 0.42 0.42 0.29 0.29 -0.02 -0.09
B3PW 2.58 2.82 1.75 1.77 1.57 1.55 1.30 1.22
On the other side, when considering the B3PW functional, all the compounds are described as
insulating systems albeit at the price of a too wide separation between valence and conduction
bands. This difference could be related to the different level of theory adopted for the calculation,
and in particular to:
a) the different diffuseness of the atomic basis set employed with PBE and B3PW.
b) the inclusion of exchange energy in the formulation of the functional, since it is completely
neglected in PBE while included as 20% of the exact HF estimate in B3PW.
c) the inadequacy of an “all-electron” description when heavy atoms are considered, since
relativistic corrections are customarily included only in pseudo-potential approaches
The issue related to point a) could be of concern for all the considered compounds. Considering the
optimized structures (Table 2.3) it can be seen how PBE is almost always superior to B3PW in
reproducing the lattice parameters and bond distances for the various compounds. The most
contracted basis set, along with the PBE functional, is the best choice when dealing with the most
diffuse atoms, As and Sb. For FeSb2 in particular, this is the only case where, after the optimization
procedure, the Fe-Sb axial bond is shorter than the equatorial one, as found experimentally.
16
Despite the aforementioned geometrical differences, negligible changes have been found for the
band gap values (Table 2.5) as well as for the DOS pictures (not shown).
The percentage of Hartree-Fock exact exchange, %HFex, added to the functional is instead found to
have a much greater impact on the separation between valence and conduction band. The impact of
this quantity on the electronic structure of a given compound is not suprising. It is not uncommon to
screen different %HFex values, in conjunction with a given DFT functional, in order to obtain the
best description of a given property of a compound. FeSb2 (experimental geometry) has been here
considered as the reference for this kind of analysis. Indeed an increase of the %HFex included in
the functional is found to enhance the separation between occupied and empty states (Figure 2.7 and
Figure 2.8). In particular, lowering the %HFex to 5% in B3PW, while raising it to the same amount
in PBE, results in band gaps of 0.22eV and 0.14eV respectively, in good agreement with the
experimental findings.
Figure 2.7 Density Of States (DOS) for FeSb2 at the experimental geometry evaluated with the PBE (left) and B3PW (right) functionals (TZ basis set) and considering different amount of %HFex (Top: 10%; Bottom: 5%). For each structure, the total and the atom-projected DOS (pDOS) are shown.
17
Figure 2.8 Bandwidth for FeSb2 at the experimental geometry evaluated with the PBE (left) and B3PW (right) functionals and considering different amount of %HFex (Top: 10%; Bottom: 5%).
However, with respect to what already obtained with the original formulation of PBE and B3PW,
no significant differences can be highlighted in how the atomic orbitals are combined to give the
resulting bonding scheme (see next paragraph).
Finally, FeSb2 has also been considered as the reference to evaluate the effects of an atoms
description based on pseudo-potentials, in order to take into account relativistic effects not
accounted for in the all-electron description. Whether they’re not particularly important in this
system or their effect is not affecting the features being considered, the final outcome is that even in
this case no differences can be highlighted (Figure 2.9). Even the band gap dependency on the
%HFex value is as observed before (0.11eV and 0.13eV for PBE and B3PW, respectively, adopting
%HFex = 5).
In conclusion, concerning orbital-based analyses, almost indistinguishable results are obtained
whether the theoretical or the experimental structure is adopted and for the different levels of
18
theory. In order to adopt the same level of theory for all the compounds while avoiding the wrong
Fe-Sb bonding distances in optimized FeSb2, the results obtained from the experimental TZ/PBE
description will be discussed in the next paragraph. However, the optimized TZ/PBE description for
FeS2 and FeP2 will also be taken into account to provide a more direct comparison with the results
obtained along a gedanken “reaction” from M to L for FeS2, for which optimized geometries were
used.
Figure 2.9 Density Of States (DOS) for FeSb2 the experimental geometry evaluated with the PBE (left) and B3PW (right) functionals (PP basis set). For each structure, the total and the atom-projected DOS (pDOS) are shown.
4.2. Electronic structure: orbital-based analysis
Figure 2.10 introduces further detail by showing the contributions from individual Fe d orbitals to
the DOS of FeS2 and FeP2. The picture arising from these projected DOS for FeS2 (Figure 2.10 left)
is qualitatively consistent with the view that LFT has on the marcasitic distortion from the perfect
octahedral field: dx2-y2 and dz2 orbitals (eg-like) give rise to bonding and antibonding states through
combination with the S px, py and pz orbitals. Conversely, dxy, dxz and dyz (t2g-like) behave as
nonbonding orbitals, with dxy slightly shifted to higher energies. It is thus possible to distinguish
three regions in such orbital projected DOS (Figure 2.10). The first, denoted as B, is dominated by
19
the bonding interactions and it is peaked at ~ -4eV with respect to the FL at 0 eV. The second,
named as AB and peaked at ~ +2eV, is the anti-bonding counterpart, while the third, that we call
NB, is dominated by nonbonding interactions and peaks at ~ -1eV. The small DOS that can be
found in the B region due to the nonbonding orbitals, as well as the DOS attributed to dx2-y2 and dz2
in the NB region, are likely to be caused by the marcasitic distortion.
Figure 2.10 FeX2 Individual Atomic Orbitals projected DOS (pDOS) for the optimized FeS2 (left) and FeP2 (right) structures. Fe d orbital states (top); X p orbital states: X-axial (middle) and X-equatorial (bottom). The blue, red and green boxes on the abscissa highlight the Bonding (B), Anti-Bonding (AB) and Non Bonding (NB) energy regions, respectively (see text).
20
Mulliken orbital populations (Table 2.6) condense the information from the projected DOSs into
numeric population values for each orbital and take implicitly into account the position of the FL
(orbital contributions are integrated up to the FL). These population values necessarily comply with
the qualitative picture sketched above while providing quantitative detail. The t2g-like states have a
larger population than the eg-like orbitals, though still far from two, which is the value one would
expect from a fully ionic picture (Fe+2) and from their ideal “non bonding” nature. The eg-like
orbitals are found to be mostly half-filled instead of being empty as required by the LFT.
According to the Mulliken analysis, the resulting total d-population on Fe is 6.78 e-, a value in
between a purely covalent and a purely ionic picture.
Table 2.6 Fe 3d orbitals’ Mulliken populations for the experimental FeX2 (X=S, P, As, Sb) structures.
Fe 3d orbital FeS2 FeP2 FeAs2 FeSb2 yz 1.61 1.64 1.64 1.64 xz 1.61 1.64 1.64 1.64 xy 1.77 1.13 1.13 1.13
x2-y
2 0.85 1.30 1.32 1.34
z2 0.95 1.09 1.09 1.13
The B, AB and NB regions can also be identified for FeP2 (Figure 2.10 right). However, at variance
with FeS2, the dxy orbital in FeP2 no longer contributes in the NB region, while it does so in both the
B and AB regions. According to the fully ionic picture,12 two more electrons (Fe+4) are withdrawn
from iron in FeP2, with the dxy orbital being emptied and raised above the Fermi level. In reality, the
population of the dxy orbital (1.13 e-, Table 2.6) is indeed highly decreased with respect to that in
FeS2 (1.77 e-, Table 2.6), however, its value close to one indicates that this orbital is involved in
bonding (figure 10 right) and that it is contributing to the DOSs both below and above the FL. At
variance with the ionic picture, the total Mulliken population of the iron d orbitals remains almost
unchanged (6.80 e-) to the value it has in FeS2. The almost constancy in the d orbitals population is
21
the result of the mentioned decrease of the dxy population (from 1.77 to 1.13 e-) being almost
compensated for by a significant increase of the dx2-y2 population (from 0.85 to 1.30 e-). In FeP2
both dx2-y2 and dxy become involved in the bonding in the equatorial plane, but the two orbitals
exchange their roles with respect to FeS2: dx2-y2 tends to become non bonding, while dxy acquires a
significant bonding character. This view becomes particularly clear and it is rigorously proven when
bonding analysis is performed in terms of observables (see next paragraph).
The question arises whether the observed change in the electronic structure of iron is primarily the
result of the compression along c on going from M to L or is rather due to the reduction of two
valence electrons when the 16-group replace the 15-group elements in TX2. To elucidate this point,
an imaginary “reaction” from M to L for FeS2 has been considered, by retaining the Pnnm space
group and taking the optimized FeS2 and FeP2 unit cells as the starting and final points of such
transformation. The lattice parameters of three intermediate structures S1, S2 and S3 (Table 2.7)
have been obtained as equidistant points of a linear variation of the cell edge values along the M to
L reaction path. The fractional coordinates in Table 2.7 are the result of the supposed linear
variation of Fe-X distances along the transformation. Since FeS2 and FeP2 have an equal ratio of the
axial to the equatorial Fe-X distances, the ratio of such distances remains automatically constant
(Table 2.7) along the reaction path. Differently from the changes observed between FeS2 and FeP2,
the d-orbital populations (Table 2.7) are marginally affected by the only geometrical variation
considered in the FeS2 M to L transformation. Moreover changes take place almost only during the
last step of the transformations (i.e. from S3 to L). Though much smaller, the observed changes are
qualitatively similar to those found between FeS2 and FeP2, with the population of dx2-y2 increasing
from 0.86 to 1.02 e- and that of dxy decreasing from 1.77 to 1.66 e-. The total electron d-population
remains almost unchanged, around to 6.81 ± 0.3 e- along the whole reaction path.
22
Table 2.7 FeS2 marcasite (M) to löllingite (L) transformation. S1, S2 and S3 are the intermediate structures obtained as equidistant points of a linear variation of the cell edge values along the M to L reaction path. Lattice parameters a, b, and c are given in Ångstrom; x and y are the fractional coordinates of the S atom; ax. and eq. identify the Fe-S equatorial and axial interactions and their bond lengths are given in Ångstrom; α, given in degrees, is the X-Fe-X “compression” angle; PMull are the Mulliken’s populations for the various Fe 3d orbitals.
M S1 S2 S3 L
a (Å) 4.4426 4.5759 4.7092 4.8425 4.9758 b (Å) 5.4114 5.4717 5.5321 5.5925 5.6529 c (Å) 3.3956 3.2275 3.0594 2.8914 2.7233
x 0.2051 0.1913 0.1813 0.1748 0.1703 y 0.3763 0.3757 0.3740 0.3711 0.3676
Fe-S ax. (Å) 2.2311 2.2343 2.2382 2.2414 2.2444 Fe-S eq. (Å) 2.2464 2.2499 2.2535 2.2560 2.2594
δ (°) 98.2 91.7 85.5 79.7 74.1 PMull (Fe-dyz) 1.60 1.60 1.60 1.58 1.57 PMull (Fe-dxz) 1.60 1.60 1.60 1.58 1.57 PMull (Fe-dxy) 1.77 1.78 1.78 1.74 1.66
PMull (Fe-dx2-y2) 0.86 0.86 0.88 0.94 1.02 PMull (Fe-dz2) 0.97 0.96 0.96 0.98 1.02
It is worth noting that the DOSs portraits displayed in Figure 2.11 reveal a qualitative difference
between the M and L electron structures of FeS2 which was hidden in the analysis of the d-orbital
population: while the dx2-y2 and dxy contributions show only a marginal overlap below the Fermi
level in the M structure, they share instead a similar energy range in the L geometry, denoting an
initial tendency to be more similarly involved in bonding. However, such tendency realizes in
practice only when S is substituted for P.
4.3. Electronic structure: observable-based analysis
According to the QTAIM theory, 29 the structure of a compound can be revealed through the
topological analysis of ρ by studying its critical points (cps); in particular those cps called bond
critical points (bcps) which tell us how atoms get bonded to one another.
23
Figure 2.11 Individual Atomic Orbitals projected DOS (pDOS) of the Fe d orbital states in the M, S1, S2, S3 and L structures of FeS2 (from top to bottom). The black arrows highlight the splitting of the dxy orbital.
24
At bcps, the electron density attains its minimum value along the bond path, the line of maximum
electron density with respect to any arbitrary displacement which links the two bonded atoms.29 For
each of the studied systems, either considering the experimental or theoretical structures, the same
number, type and Wyckoff position of cps were found Three bcps were identified, B1, B2 and B3,
related respectively to the axial and equatorial Fe-X interaction, and to the short X-X interaction
(X = S: 2.154Å; X = P: 2.2365Å; X = As: 2.4915Å, X = Sb: 2.8941Å) (Figure 2.12).
Figure 2.12 B1, B2 and B3 type bond critical points (bcps) in FeX2. Fe atom is represented by a blue sphere, X atoms by yellow and bcps by small black spheres. Two X2 dumbbells have also been added to highlight the presence of an interaction (B3 bcp) between the atoms constituting the dumbbell but not between the atoms of different dumbbells (dashed lines).
It is worth noting that no bcp, hence no direct X⋅⋅⋅X interaction (X = S: 3.3871Å; X = P: 2.7230Å;
X = As: 2.8822Å, X = Sb: 3.19733Å), was found between the diatomic anions (X2) in the direction
along c (Figure 2.12), although Papoian and Hoffmann14 have hypothesized its occurrence to fulfill
the 18-electron counting rule in FeSb2. As for what concerns the topological parameters at the
various cps (Table 2.8), few differences can be highlighted between the experimental and the
optimized structures, the only one worth mentioning being the change in sign of the Laplacian, ∇2ρ,
for the S-S bcp.
25
Table 2.8 FeX2 (X=S,P) bcp properties for the experimental (exp.) and optimized (opt.) structures. B1, B2 and B3 denotes the axial, the equatorial Fe-X interaction, and the short X-X interaction, respectively; dFe-X and dX-X are the Fe-X and X-X distances (in Ångstrom); AIL is the Atomic Interaction Line29 (called bond path at equilibrium) and AILX-bcp (AILFe-bcp) are the distances from X (Fe) atom to bcp, on moving along the AIL; ρ and ∇2ρ are the electron density and the Laplacian of the electron density at the bcp, in e/Å3 and e/Å5 respectively; ε is the ellipticity at the bcp. The values in parentheses refer to the optimized structures.
FeS2 FeP2 FeAs2 FeSb2
B1
dFe-X (Å) 2.2359
(2.2311) 2.2484
(2.2444) 2.3620
(2.3625) 2.5746
(2.6216)
AILFe-bcp (Å) 1.0095
(1.0092) 1.0503
(1.0495) 1.0864
(1.0873) 1.1702
(1.1811)
AILX-bcp (Å) 1.2266
(1.2220) 1.1982
(1.1951) 1.2759
(1.2755) 1.4049
(1.4409)
ρ (e/Å3) 0.568
(0.575) 0.592
(0.596) 0.507
(0.507) 0.415
(0.387)
∇2ρ (e/Å5) 5.151
(5.117) 2.767
(2.792) 2.320
(2.300) 1.226
(1.185)
ε 0.034
(0.032) 0.009
(0.004) 0.002
(0.001) 0.024
(0.027)
B2
dFe-X (Å) 2.2539
(2.2464) 2.2637
(2.2594) 2.3874
(2.3813) 2.5965
(2.5915)
AILFe-bcp (Å) 1.0201
(1.0187) 1.0881
(1.0872) 1.1299
(1.1279) 1.2129
(1.2098)
AILX-bcp (Å) 1.2423
(1.2363) 1.1764
(1.1729) 1.2580
(1.2539) 1.3848
(1.3825)
ρ (e/Å3) 0.548
(0.558) 0.563
(0.567) 0.478
(0.483) 0.388
(0.390)
∇2ρ (e/Å5) 5.012
(5.030) 1.928
(1.918) 1.718
(1.728) 0.995
(1.042)
ε 0.011
(0.007) 0.024
(0.024) 0.034
(0.034) 0.075
(0.056)
B3
dX-X (Å) 2.2154
(2.2613) 2.2365
(2.2610) 2.4915
(2.5135) 2.8941
(2.8930)
AILX-bcp (Å) 1.1077
(1.1307) 1.1184
(1.1307) 1.2458
(1.2568) 1.4471
(1.4466)
ρ (e/Å3) 0.743
(0.687) 0.742
(0.714) 0.529
(0.512) 0.371
(0.372)
∇2ρ (e/Å5) -0.256 (0.109)
-2.735 (-2.438)
-0.543 (-0.453)
-0.170 (-0.184)
ε 0.032
(0.029) 0.022
(0.022) 0.008 0.008
0.012 (0.002)
26
By optimizing the geometry, the S-S bond lengthens and the Laplacian becomes slightly positive
instead of being negative and very close to zero as in the experimental geometry structure.
However, although positive (0.109 eÅ-5), it is much less so than the value (2.588 eÅ-5, Table 2.9)
one would obtain by placing two non interacting S atoms at the same internuclear distance (Fig.
2.13). This observation implies that the S-S interaction, compared to the non-bonding case,
maintains a (tendency to a) shared-shell (covalent) bond nature even in the optimized structure
despite the Laplacian being positive at the bcp.
Table 2.9 FeX2 (X=S,P) : Laplacian values at the X-X bcp in the crystal and those in the free atom X, multiplied by two, at the bcp distance from the nucleus in the crystal. ∆∇2ρ are the differences between ∇2ρ values at X-X bcp in the crystal (at optimized or experimental geometry) and those, multiplied by two, in the free atom X at the bcp distance from the nucleus in the crystal.
∇2ρ (e/Å5) ∆∇2ρ (e/Å5) crystal (exp) crystal (opt) free atom (2x) exp opt
S-S -0.256 0.109 2.588 -2.844 -2.479 P-P -2.735 -2.438 1.050 -3.785 -3.488
As for what concerns the Fe-X interactions, ∇2ρ is found to be positive in all cases. The bcp is
located at about 1 Å from the iron and at about 1.2 Ǻ from the S or P atoms, that is at distances
which fully fall (Figure 2.13) in the depletion region of the M quantum shell of all these atoms
when isolated (and neutral).
The resulting positive Laplacian at the bcp is therefore not surprising, the observed values being
about 5 and 7% (ax, eq, respectively; FeS2) and 26 and 37% (ax, eq, respectively; FeP2) smaller
than the sum of the isolated atom contributions. The higher percentage decrease found for FeP2 is
likely a sign of an enhanced covalency in this compound with respect to FeS2. Such finding is also
corroborated by the smaller Bader net charge found on iron for FeP2 (+0.25e- rather than +0.67e- in
FeS2, Table 2.10).
27
Figure 2.13 Laplacian profiles along the X-X (top left: X = S; bottom left: X = P; one atom is placed at the origin) and Fe-Xeq (top right: X = S; bottom right: X = P; the Fe atom is placed at the origin) interaction lines. The profiles portrayed refers to the crystal (solid blue), to the isolated atoms (Fe: solid black; X: dashed black) and to the sum of the isolated atoms (solid green) Laplacian densities. The black dots mark the position of the bond critical point (bcp) in the crystal. On the x axes is reported the distance from the origin.
Table 2.10 Bader’s electron populations, N(Ω), and net charges, q(Ω) for the experimental structures.
Atom Ω FeS2 FeP2 FeAs2 FeSb2
Fe N(Ω) 25.33 25.75 25.85 26.17 q(Ω) 0.67 0.25 0.15 -0.17
X N(Ω) 16.33 15.13 33.07 50.91 q(Ω) -0.33 -0.13 -0.07 +0.09
The small departure of the bcp density from that found with the non-interacting model is in line
with a dative bonding description for FeS2. Indeed, the electron density at the Fe-X bcp is as much
28
as 20% smaller than the sum of the isolated atom densities in FeS2, whereas, in FeP2 , in agreement
with the enhanced shared character of the Fe-X interaction, it is larger than this sum by about 1-4%.
Further interesting insight into the relationships between d-orbital populations, metal-ligand
interactions and formal number of electrons on the metal in the studied compounds is obtained by
considering the Laplacian density topology.29 As it has been repeatedly shown in a number of
papers,30,31,32,33,34 the Laplacian distribution reflects the deviation from spherical symmetry which
results from an incomplete filling of the valence shell orbitals of an atom. By combining a study of
the Laplacian topology with an orbital-based analysis, light is shed on the correspondence between
the portrait of an observable in a given system and the expectations of the orbital model and of the
LFT in an idealized representation of this same system.30,31,32 In a perfect octahedral field, the total
t2g orbital electron density has eight lobes pointing along the diagonal of a cube while the density
associated with the eg orbitals has six lobes pointing towards the faces of a cube.35 If the t2g orbitals
are filled and the eg are empty, as it occurs in the Cr (t2g)6 configuration,31 the resulting Laplacian
distribution has a cubic shape with density accumulations along the diagonals of the cubes and
holes(depletions) at the center of the faces. It has also been shown36,37 that the donor-acceptor
interaction between a metal atom and the surrounding ligands may be often interpreted in terms of a
so-called ‘key-lock’ mechanism, where a -∇2ρ lump on the ligand is aligned to, and matches with, a
-∇2ρ hole on the metal. This for instance occurs31 in the case of the prototypical Cr(CO)6 metal-
ligand complex whose key-lock arrangement of charge concentrations and depletions is opposite to
the one usually found for AYn complexes,33 in which A is a main group element. In this latter case,
the charge concentrations in the Valence Shell Charge Concentration (VSCC)37 region of A are
directed towards the ligands and neither lie in-between two neighboring ligands nor are oppositely
located (LOCC, Ligand-Opposed Charge Concentrations).33 These two contrasting mechanisms
denote the limiting situations of bonding in the two cases, a closed-shell dative interaction for
29
Cr(CO)6 and a shared-shell, covalent or covalent-polar interaction for AYn. Our systems differ from
Cr(CO)6 under many aspects. First, a distortion from perfect octahedral symmetry occurs, which
leads to partial filling of all orbitals even for a d-electron population on the metal lower than six.
Secondly, the number of d-electrons on the metal is not exactly six and its estimated value depends
on the adopted model or on the chosen charge partitioning scheme. We already noted that in the
fully ionic model such value amounts to six for FeS2 and four for FeP2, while we found it to be
equal to 6.94 e- and 6.92 e-, respectively, in the Mulliken’s approach and to 7.33 e- and 7.74 e-
according to Bader’s partitioning. b Finally, the dative and covalent bonding schemes enter in
competition in these systems, with the second scheme largely enhancing its weight in FeP2
compared to FeS2.
In the ∇2ρ contour maps for FeS2 (Figure 14 left, top and middle), a t2g6-like octahedral VSCC can
be recognized for the iron atom, with the regions of charge depletion (or, more precisely, of less
charge accumulation)c being approximately aligned to the S regions of charge accumulation in a
pseudo key-lock structure.31 This is made clear in Fig. 2.15 left (top and in particular bottom) where
one notices that the -∇2ρ maxima on Fe bisect the angles formed by the -∇2ρ maxima on S pointing
towards the iron nucleus and this same nucleus as central vertex of these angles. A closer exam of
Figure 2.14 (left, middle) and Fig. 2.14 (left, bottom) reveals, however, that in the equatorial plane a
slight but noticeable misalignment of the charge concentration with respect to the charge depletion
regions also occurs. This can easily be explained by the geometrical distortions occurring in this
b In the case of Bader’s partitioning, individual orbital populations are not defined. The given
estimate of d orbital population, according to this partition, assumes that the iron atom has a frozen Ar core and that its 4s2 electrons have been entirely transferred to the ligands.
c With respect to the ideal t2g6 octahedral case, the regions associated with the eg orbital densities,
rather than being regions of charge depletion, are, for our systems, regions of decreased charge concentration only. As we said before, the population of the Fe d orbitals exceeds 6 e- in FeS2 (and FeP2) and the geometrical arrangement of the ligands around the Fe differs from the ideal octahedral case. Both these facts have an impact on the distribution of d electrons density around the Fe atom and hence on the portrait of the total ∇2ρ distribution.
plane (the δ angle significantly deviates from 90
of the known relationships between the properties
the S charge concentrations directed towards
diminish their overlap with the neighboring concentration associated to the smaller S
(1800-δ, see Table 2.2). In such a way, the repulsion between the charge concentrati
neighboring pairs of S atoms is decreased, while that of the farther neighboring pairs is enhanced,
making these repulsions more alike to each other.
Figure 2.14 ∇2ρ maps for the optimized FeSmark regions of charge accumulation (negative regions of charge depletion (positive the axial Fe-X (top), equatorial Feshown. The pictures in the middle identify the considered planes.
angle significantly deviates from 900, Table 2.2) and, quite interestingly, also in terms
of the known relationships between the properties of the VSCC and the VSEPR model
ions directed towards the Fe atom rotate in the equatorial plane so as to
diminish their overlap with the neighboring concentration associated to the smaller S
). In such a way, the repulsion between the charge concentrati
neighboring pairs of S atoms is decreased, while that of the farther neighboring pairs is enhanced,
making these repulsions more alike to each other.
maps for the optimized FeS2 (left) and FeP2 (right) structures. Solid mark regions of charge accumulation (negative ∇2ρ) while dotted red lines mark regions of charge depletion (positive ∇2ρ). For each structure, the plane defined by
X (top), equatorial Fe-X (middle) and X-X (bottom) interactions is own. The pictures in the middle identify the considered planes.
30
2) and, quite interestingly, also in terms
of the VSCC and the VSEPR model.38,37 Namely,
the Fe atom rotate in the equatorial plane so as to
diminish their overlap with the neighboring concentration associated to the smaller S-Fe-S’ angle
). In such a way, the repulsion between the charge concentrations of the closer
neighboring pairs of S atoms is decreased, while that of the farther neighboring pairs is enhanced,
(right) structures. Solid blue line ) while dotted red lines mark
). For each structure, the plane defined by X (bottom) interactions is
own. The pictures in the middle identify the considered planes.
31
Figure 2.15 ∇2ρ topology around the Fe atom and along the Fe-X interactions for the experimental structures. The large blue, yellow, purple, green and red spheres represent the Fe, S, P, As and Sb atoms respectively. The small black spheres mark the position of the bcp while the small blue spheres define the position of the ∇2 ρ minima (region of charge accumulation) found within Fe and VSCC. The inset shows an enlargement of the charge accumulations spatial distribution. The values of ρ (in e/Å3) and ∇2 ρ (in e/Å5) for the bcp and ∇2 ρ minima are reported in normal and italics respectively.
32
The misalignment of the S concentrations with respect to the (S-Fe)eq axis is further evidenced in
Figure. 2.16, which shows the four -∇2ρ VSCC maxima around each S atom. Their location nicely
complies with a sp3-like hybridization of S. The mentioned misalignment enables those two
concentrations of S pointing towards two equatorial iron atoms not only to stay further apart than
implied by the Feeq-S-Feeq angle (97.94°, equal to the S-Fe-S δ angle, Table 2.2 and Figure 2.3), but
to even surpass the sp3 tetrahedral value by about six degree (115.4°). In such a way, the distance
and the repulsion among the charge concentrations involved in the dative-like bonds (those with
more “lone-pair”, or more precisely, non-shared pair character)d are, respectively, maximized and
minimized. Likewise, also the two S charge concentrations pointing towards an equatorial and an
axial Fe atom, though slightly decreasing their angle with respect to the FeeqSFeax angle, still retain
an angle of 117.1°, significantly larger than the tetrahedral value. Instead, any pair of S
concentrations pointing toward the bonded sulphur atom S’ and either one axial or one equatorial Fe
atom form a much narrower FeSS’ angle of about 101°, since the S charge concentration directed
towards S’ has more a bonding rather than a lone-pair like character. In conclusion, the
misalignment of the S charge concentrations with the relative charge depletions on Fe has clearly a
twofold origin. On one hand it allows an equalization of the repulsion between lone-pair like
concentrations of different S atoms (Figure 2.14 left, middle and Figure 2.15, left) and on the other
hand it enables the minimization of the repulsion between the lone-pair like concentrations on each
S atom (Figure 2.16 left).
d Clearly, when referring to bonding or lone-pair (“non-shared” pair) character of the charge
concentrations, it is not to be intended that their associated regions will include a full electron pair: it simply refers to the more or less shared nature of the fraction of an electron pair enclosed in that region.
33
Figure 2.16 ∇2ρ topology of the X atom for the optimized FeS2 (left) and FeP2 (right) structures. The large blue, yellow and purple spheres represent the Fe, S and P atoms respectively. The small black spheres mark the position of the bcp while the small green spheres define the position of the ∇2ρ minima (region of charge accumulation) found within X VSCC respectively. The values of ρ (in e/Å3) and of ∇2ρ (in e/Å5) at bcp and at -∇2ρ maxima are reported in normal and italics respectively.
We discuss now the FeP2 system. In the ∇2ρ contour maps for FeP2 (Figure 2.10 right, top and
middle), the t2g6-like octahedral VSCC is hardly recognized for the iron atom, as opposed to what is
the case for FeS2. The regions of charge depletion and charge accumulation are now very
marginally differentiated, especially so in the plane of the equatorial (Figure 2.15) Fe-P bonds. In
summary, the key-lock pattern of charge concentrations and depletions, typical of metal-ligand
interactions, seems to be lost. Although the close exam of the -∇2ρ topology around Fe still reveals
the presence of eight charge concentrations, the Laplacian distribution has no longer an almost
regular cubic shape, but has acquired that of a cuboid (Figure 2.15) with one side almost half as
long, 0.24 Å, than the other two sides, 0.42 Å (for the sake of comparison the corresponding figures
in FeS2 are 0.32 and 0.38 Å)
As a consequence, the Fe charge concentrations approach each other in pairs (Figure 2.15) and,
rather than bisecting, from above and below, the Peq-Fe-Peq angles, result almost aligned, from
34
above and below, to the Fe-Peq bonds (Figure 2.15). This bonding situation approaches that usually
found for AYn complexes where A is a main group element. This is in nice compliance with the
increased covalent character found for the Fe-X interaction when P substitutes S. The change in the
Laplacian distribution around the iron atom also agrees with the observed decrease of the dxy and
parallel increase of the dx2-y2 (and also dz2) population on moving from X=S to X=P (see previous
paragraph). It is worth noting that the observed rotation of the iron VSCC maxima with respect to
the ligands is primarily due to the different electronic nature of the S and P “ligands”, rather than a
result of the change of geometry from M to L structure. Indeed, when FeS2 with the L structure is
considered, the location of the eight VSCC maxima of iron turns out hardly to be affected, despite
noticeable changes in the values of ρ and ∇2ρ at such maxima.
For FeP2, the shared shell character of the P-P interaction is clearly visible (Figure 2.14 right
bottom) and, at variance with FeS2, the separate charge concentration maxima on P pointing
towards the iron atoms in the equatorial plane seem to have merged into a single maximum (Figure
2.14 right middle). A closer exam of the topology of -∇2ρ shows that it is however still possible to
distinguish four maxima also for FeP2 though largely flattened (Figure 2.16, right) with respect to
FeS2. Despite the fact that the FeeqXFeeq angle is highly narrowed from 97.9° for X=S to a value of
only 73.8° degrees for X=P, the charge concentrations pointing toward Feeq are aligned along the
Fe-P axes. There are two concurring reasons for such a behavior. Due to their increased bonding-
nature, the charge concentrations in FeP2 are no longer forced to stay as far apart as possible as it
occurs in FeS2, despite the very small FeeqPFeeq angle. The enhanced bonding character of such
concentrations is also revealed by their largely increased proximity to the B2 bcp, 0.36 Å, to be
compared to a value of 0.55 Å in FeS2. A second clear rationale of why the two VSCC maxima on
the P atoms pointing towards the Feeq atoms get closer and almost merge into a unique flattened
maximum is that the repulsion with the corresponding charge concentration maxima of the
35
neighboring P atoms along the c axis is decreased this way. In FeP2 the cell is significantly and
visibly shrunk along the c axis; if the VSCC maxima on each P pointing towards the Feeq atoms
were not flattening and not rotating towards each other, they would experience a largely enhanced
repulsion with respect to the case of FeS2 (compare in Figure 2.14, middle left and right contour
plots).
5. Conclusions
This paper disentangles the peculiar and subtle interplay between geometry, electronic structure and
bonding interactions in two marcasites and löllingites prototypical compounds. This goal was in
part achieved by performing DFT periodic calculations on FeS2 and FeP2 followed by a comparison
of the predictions of ligand field theory against the description given by an orbital-based analysis
and by the composition of bonding and antibonding states close to the Fermi level. The most
illuminating and comprehensive picture was yet obtained through the analysis of the topology of
two observables, the electron density and the electron density Laplacian, using the QTAIM
approach. In particular, qualitative and quantitative information on the valence electron distribution
anisotropy of the iron is obtained in terms of the topology of the Laplacian density, and especially
so when compared to the portrait of the corresponding topology for an idealized octahedral t2g6
model of the iron metal atom.
So far, in the literature, the differences between the bonding interactions in the M and L type of
compounds were essentially ascribed to the extent of charge transfer from the transition metal to its
linked anions and to the resulting modifications of the metal d-electrons asphericity. The charge
transfer is instead found to be quite small and comparable in both compounds, in contrast to the
expectations of the LFT, but in accord with electronegativity scales. Indeed, what mainly
distinguishes the two compounds is the kind of anisotropy of the electron density around the iron
36
atom and, therefore, in how the metal d orbitals are distinctively populated. The cubic arrangement
of charge concentrations around the iron atom in FeS2 transforms to a cuboid shape in FeP2, with
charge concentrations being directed towards the X anions rather than in between them in the
equatorial plane. In M compounds, electrons are distributed preferentially according to the dictate of
LFT for a (distorted) octahedral environment (though a significant electron population is also found
in the eg orbitals, x2-y
2 and z2) whereas in the L systems the x2-y
2 and xy orbitals interchange their
roles. Such different charge concentrations or orbital’s population distributions are, rather than the
result of the geometry change from M to L compounds, essentially driven by the nature of the anion
(or ligand). Pnicogens, having one less electron, are able to bond to the iron with much higher
covalent character than chalcogens do. Indeed, when FeS2 is forced to adopt the L geometrical
structure of FeP2, although the changes occur in the expected direction from M to L compounds,
only a very partial electron density rearrangement around the iron can be observed.
As clearly revealed by the greater uniformity of charge concentrations around X (Laplacian density
topology), Fe-X and X-X bonds are significantly more alike to each other when X is a a pnicogen
than when it is a chalcogen. The enhanced covalent character of the Fe-X bonds when X is a
pnicogen is also evident from all other kinds of analyses that were performed. On going from FeS2
to FeP2 a noteworthy change in the d orbital populations of iron, in the energy location and orbital
composition of the DOS profiles, and finally in the electron density topology, was observed. These
changes correspond to an increased (decreased) weight of X p orbital contributions to the bonding
(antibonding) states close to the Fermi level, a diminished charge transfer and a clearly enhanced
covalent character of Fe-X and X-X bonds.
To summarize, the LFT predictions are better recovered in the case of FeS2 than they are for FeP2
because in the former compound the anion is of a nature that is closer to the assumption of a
37
negative ligand inherent to the LFT while in the latter the anion departs more significantly from
such assumption being more prone to covalent interactions.
Careful X-ray diffraction experimental studies, aimed at obtaining charge-density quality data of a
number of M (X=S) and L (X=P, As, Sb) crystals, are currently underway in the laboratory of the
Chemistry department of the Aarhus University. Obtaining a meaningful charge density for these
systems is quite challenging in view of their relatively high symmetry and, more importantly, of the
small ratio of the valence to core electrons. Both these factors lead to known difficulties in the
multipolar model refinement of diffraction data. When available, experimental charge densities and
density Laplacian distributions will be analysed within the same topological approaches used in the
present study, enabling a direct and fruitful comparison with the theoretical investigation. Two main
scopes may be envisaged through this coming analysis. On the one hand, to test whether the
interesting description and conclusions put forth by the theoretical approach, like the change in d
orbital populations and in the metal VSCC shape on passing from M to L, are sufficiently stable
against the ambiguities and freedom inherent to the treatment of the experimental data. On the other
hand to investigate whether crystal defects, synthesis conditions, presence of impurities, etc. could
play some role in the delicate electronic balance leading to the peculiar metal asphericities and
metal-ligand bonding features in the M and L compounds.
38
Chapter 3
Hypervalency in K2SO4
1. Introduction
A cornerstone in chemistry is the ability to predict how and why bonds are formed between atoms
leading to the rich variety of compounds that constitute the world we live in. Understanding the
principles behind chemical bonding is therefore crucial for any chemist. The Lewis theory is one of
the most basic and widespread theories on chemical bonding and it typically appears in introductory
chemistry text books.39 However, cases can be found for which the octet rule, and thus this model,
is violated because the central atom in a molecule has more than eight electrons in its valence shell.
Such compounds are called hypervalent. The sulfate ion, SO42-, is an example of an entity that is
hypervalent according to the Lewis model. The hypervalency results from the introduction of
double bonds in the structure leading to a count of 12 valence electrons for sulfur as indicated in
Figure 3.1. The idea of double bonds and thereby hypervalency is supported by the short S-O bond
lengths generally found from experiment and theory. An alternative description of the sulfate ion
involves only single bonds and this avoids the issue of hypervalency (Figure 3.1). However, this
description does not fulfill the Lewis requirement of minimizing the formal charges. The
hypervalent picture is expected to involve the presence of several resonance forms as opposed to the
single bond case. Furthermore, the number of lone pairs on the oxygen atoms differs in the two
cases. This means that a characterization of the bonding and the number of lone-pairs in the sulfate
group can give the information concerning which picture is more correct. However, since both in
the hypervalent and in the single-bonded situation (Figure 3.1) all oxygen atoms would anyhow
reflect the resonant or “static” occurrence of three lone pairs in their valence shell, only the form
39
and magnitude, rather than the number, of lone pair charge density concentrations should be
affected.
Figure 3.1 Hypervalent (left) and single-bonded (right) depiction of the bonding situation in a sulfate ion, SO4
2-. Only one of the resonance forms is drawn for the hypervalent case. Formal charges are indicated in brackets.
Several studies, both experimental and theoretical, have focused on S-O interactions and the
‘hypervalent’ character of sulfur.40,41,42 One of the arguments in favor of the hypervalent description
is the inclusion of the 3d-orbitals on sulfur. In this picture the d orbitals should be capable of
accommodating the additional valence electrons on sulfur. Theoretical calculations have proven the
necessity to include the 3d orbitals in the sulfur basis sets used for the calculations.43 However, as
pointed out by Cioslowski and Surján this could also reflect the need to include polarization
functions in the optimization in order to properly describe the effects of bond-formation. The
general conclusion made by Cioslowski and Surján for the ‘hypervalent’ molecules in their study is
that “each of the formally double S-O bonds consists of one highly polarized covalent bond and one
almost fully ionic bond.”44 This is a description where the octet rule is not violated. A similar view
on the S-O bonding is given by Dobado et al. that characterize it as ‘polar single σ-bonds mainly
characterized by electrostatic interactions’.41
This chapter reports the results obtained from the study of the electron density of a sulfate
containing crystalline compound, K2SO4 (Figure 3.2) which is chosen since it is highly ionic so that
the actual charge of the sulfate group is close to the formal charge of -2.
40
Figure 3.2 K2SO4 crystal structure.
The main aim of this study was to experimentally confirm the results found in the various
theoretical studies and to resolve the issue of the S-O bonding that has lead to the theories of
hypervalency. However, even though K2SO4 is a simple compound, it presents a substantial
challenge to experimental charge density determination due to the quite severe extinction observed
for the low order reflections.45 Accurate measurement of these reflections is important since they
carry the majority of the valence electron information. Use of a high energy, high intensity
synchrotron beam makes it possible to carry out accurate diffraction experiments on micron sized
crystals (∼30 µm) with much reduced absorption and extinction effects compared with conventional
sources.46
Mette S. Schmøkel (PhD stud., Århus university - DK) performed all the experimental work,
including data collection, reduction and modeling of the experimentally derived charge density. To
this purpose, the multipole refinement approach as implemented in the XD200647 program package
was adopted, and the programs TOPXD and XDPROP were used to perform the topological
analysis. Unless needed for the discussion, no experimental details will be reported
41
2. Computational details
The theoretical electron density has been obtained from periodic ab-initio DFT calculations
performed employing the program CRYSTAL06.16 Basis sets of def2-TZVP quality17 were used for
the various atoms along with the B3LYP functional. In order to avoid issues related to quasi-linear
dependence of the basis set, the most diffuse s, p and d Gaussian functions for the K atom were
removed.16 No modifications have been performed for the O and S basis set. The overall (s/p/d)
contraction scheme therefore was: K (8421/631/11), S (73211/51111/21), O (62111/4111/11). The
wavefunction was evaluated at the experimental geometry determined by Mette S. Schmøkel.
Additionally, two different optimization strategies were considered. Firstly, the atomic position
were optimized keeping cell edges fixed at the experimental values. Then, both atomic positions
and lattice parameters were optimized while retaining the Pnma space group symmetry. No
significant differences have been found when evaluating electronic and topological properties from
the three different periodic wavefunctions. In the following, only the results obtained from the
wavefunction evaluated at the experimental geometry will be discussed.
The topology of the density was analyzed using the beta version of the program TOPOND08.48
3. Results and Discussion
The main topological parameters are listed in Table 3.1 and a structural drawing depicting the
critical points is shown in Figure 3.3. A few discrepancies can be found between experiment and
theory. The first general considerations concern the number of critical points. For the theoretical
density the Morse relation n-b+r-c = 0 (n, b, r and c being the number of attractors, bond (BCP),
ring and cage critical points respectively) is fulfilled.49
42
Figure 3.3 S-O and O-O bonding interactions within and between the sulfate groups obtained from the theoretical electronic density. Big yellow and red spheres represent S and O atoms respectively. The small colored spheres represent the bcp position: B15 = bright green, B16 = green, B17 = orange, B18 = yellow, B19 = sky blue, S-O bcp = black. The numbers helps in identifying the (secondary) sulfate groups surrounding the central one. Some oxygen atoms have been removed from the secondary groups for the sake of clarity.
When considering the experimental electron density, a few critical points found in the theoretical
density are missing. Alternative critical points, replacing the missing ones, could not be located,
although they unavoidably exist, in order to fulfill Morse’s relation (valid also for experimental
densities derived through a multipolar moment model approach since a analytic function and its
derivatives are eventually calculated). The missing BCPs are related to the longest (more than 3 Å)
K2-O2 and O1-O3 interactions. As can be seen in Figure 3.4 the electron density and its Laplacian
are almost entirely flat in the region of these BCPs. Given the number and complexity of the
chemical interactions in K2SO4 it is likely that the apparent lack of the mentioned critical points in
the experimental density is due to numerical inaccuracies.
43
Table 3.1 Topological parameters for the different chemical bonds based on the experimental (first line) and theoretical (second line) charge density. AIL (Å) is the length of the atomic interaction line. ρ is the electron density (e/Å3), ∇2ρ the Laplacian (e/Å5) and ε the ellipticity of the bond. The labels B1 to B19 are used throughout the text to identify the various BCPs.
AILS-bcp (Å) AILbcp-O
(Å) ρ (e/Å3) ∇2ρ(e/Å5) ε
S-O1 (B14)
exp 0.591 0.889 2.017 -0.213 0.014 opt 0.590 0.891 1.943 10.221 0.004
S-O2 (B12)
exp 0.588 0.879 2.038 1.280 0.018 opt 0.585 0.882 1.988 12.355 0.004
S-O3 (B13)
exp 0.591 0.885 2.031 1.107 0.076 opt 0.588 0.889 1.957 10.817 0.004
O1-O3 (B15)
exp - - - - - opt 1.721 1.714 0.027 0.351 3.363
O1-O3 (B16)
exp - - - - - opt 1.751 1.750 0.025 0.306 1.879
O2-O2 (B17)
exp 1.671 1.671 0.037 0.363 0.229 opt 1.670 1.670 0.034 0.422 0.071
O2-O3 (B18)
exp 1.695 1.730 0.035 0.304 0.447 opt 1.696 1.710 0.028 0.353 0.138
O3-O3 (B19)
exp 1.651 1.651 0.037 0.368 0.419 opt 1.656 1.656 0.031 0.426 0.313
K1-O2 (B1)
exp 1.394 1.297 0.116 1.914 0.004 opt 1.388 1.302 0.105 2.047 0.000
K1-O1 (B2)
exp 1.469 1.412 0.086 1.275 0.189 opt 1.445 1.433 0.082 1.440 0.067
K1-O3 (B3)
exp 1.476 1.426 0.084 1.227 0.221 opt 1.453 1.444 0.078 1.380 0.085
K1-O3 (B4)
exp 1.529 1.530 0.069 0.918 0.851 opt 1.513 1.529 0.060 1.015 0.332
K1-O3 (B5)
exp 1.531 1.549 0.070 0.925 1.319 opt 1.516 1.539 0.060 1.026 0.669
K2-O1 (B6)
exp 1.392 1.312 0.117 1.961 0.073 opt 1.371 1.332 0.119 2.159 0.024
K2-O3 (B7)
exp 1.402 1.305 0.111 1.865 0.005 opt 1.395 1.313 0.103 1.978 0.014
K2-O3 (B8)
exp 1.422 1.345 0.102 1.668 0.048 opt 1.402 1.365 0.103 1.828 0.050
K2-O1 (B9)
exp 1.424 1.331 0.101 1.660 0.008 opt 1.419 1.336 0.092 1.759 0.013
K2-O2 (B10)
exp 1.466 1.409 0.085 1.338 0.297 opt 1.446 1.426 0.082 1.473 0.137
K2-O2 (B11)
exp - - - - - opt 1.549 1.557 0.050 0.873 1.051
44
This is also supported by the fact that the profiles of the Laplacian along the different interatomic
distances is very similar for experiment and theory. For all the K-O and O-O interactions the
Laplacian profiles are similar to the one shown in Figure 3.4. The BCP is located far from the
atomic valence shell charge concentrations (VSCCs) and the value of ρ is very low. This, combined
with positive values of the Laplacian at the BCP, is an indication that these interactions have the
expected closed shell nature.
Figure 3.4 Laplacian profiles along the K-O (left) and O-O (right) interaction lines in the crystal. The dotted vertical line marks the position of the bond critical point (bcp). On the x axes is reported the distance from the origin (in the left picture the K atom is placed at the origin).
In contrast to this, the profiles for the S-O interactions are different and also show some deviation
between experiment and theory (Figure 3.5). The experimental Laplacian values at the BCPs for the
three S-O interactions (Table 3.1) would indicate a different description for the three bonds. In fact,
just considering the sign of these values would indicate a shared shell interaction for S-O1 and
closed shell interactions for S-O2 and S-O3. However, the values are rather small and therefore do
not adequately assess the nature of the bonds. In contrast, the three theoretical values are similar in
magnitude and clearly positive. The reason for the discrepancies in the Laplacian values at the S-O
BCP is made clear in Figure 3.5. The Laplacian is quite steep around the position of the BCP and,
as a consequence, a small variation of the position of the BCP will change the Laplacian drastically
45
and may even change the sign of it. The overall picture of the S-O bonding obtained from the
Laplacian profiles is that charge accumulation can be observed in the region between the two types
of atoms. Yet, due to the large charge transfer between S and O atoms, this region fully belongs to
the oxygen atom basin. In Figure 3.6 a comparison is made between the Laplacian profiles for free
and bonded atoms. It is seen that the general shape of the sulfur and oxygen VSCC can still be
recognized in the molecule. However, the sulfur atomic VSSC is upon bond formation fully
attributed to the oxygen atomic basins. As a consequence, and in accordance with the different
electronegativities of the S and O atoms, the major part of the electronic charge participating in the
S-O bonding is shifted towards the oxygen. The positive value found in some cases for the
Laplacian at the BCP is simply due to the fact that the BCP is located in the region of charge
depletion of the sulfur atom.
Figure 3.5 Laplacian profiles along the three S-O interaction lines in the crystal for the theoretical and the experimental data. The dotted vertical line marks the position of the bond critical point (bcp). On the x axes is reported the distance from the origin and in each picture the S atom is placed at the origin.
46
Figure 3.6 Laplacian profiles along the S-O1 interaction lines in the crystal for the theoretical data. The profiles portrayed refers to the crystal (solid black) and to the isolated atoms (S: solid red; O: solid blue) Laplacian densities. The black dots mark the position of the bond critical point (bcp) in the crystal. On the x axes is reported the distance from the origin.
Even though the details differ the overall picture of the S-O bonding is consistent for the
experimental and theoretical densities. For all S-O bonds localized charge accumulation is observed
in the region between the two atoms which is characteristic of a shared-shell interaction.
Furthermore, the BCPs are located closer to the less-electronegative sulfur atom indicating a highly
polarized bond. Also, they are found to lie close to the Laplacian zero-contour region as seen in
Figure 3.5. This is a clear proof of the transfer of charge from sulfur to oxygen. The profile of the
Laplacian along the S-O bonds shows two local minima and a local maximum (i.e. a ‘double dip’ in
∇2ρ) in the region of the BCP. In a previous study such a feature has been taken as an indication of
covalent character of the bonding.42
This description of the S-O bond is supported by the two-dimensional contour plots and three-
dimensional isovalue surfaces of the Laplacian of the electron density of the sulfate group in
Figures 3.7 and 3.8. Four VSCC maxima and associated regions (that is their irreducible domains)
are found around sulfur directed towards the surrounding oxygen atoms and essentially belonging to
these atoms, since the -∇2ρ maxima are located in their basins. From the contour plots in Figure 3.7,
47
three charge accumulations can be envisaged within the VSCC of each oxygen atom. It is tempting
to identify these with two oxygen lone-pairs and the S-O bonded concentration. From the three-
dimensional isovalue Laplacian surfaces in Figure 3.8 it is not easy to identify localized regions of
charge accumulation for the oxygen atoms. Only for O3, and using a contour level as high as
-85e/A3 it is possible to see three distinct accumulation regions.
Figure 3.7 ∇2ρ maps cut in the O1-S-O2 (left) and (O3-S-O3) right plane of the crystal. Solid blue line mark regions of charge accumulation (negative ∇2ρ) while dotted red lines mark regions of charge depletion (positive ∇2ρ).
Figure 3.8 Three-dimensional plots of the Laplacian at contour levels of zero (left) and -85 e/Å5 (right) for the experimental density.
48
In order to more adequately identify VSCC maxima and their associated regions around the oxygen
atoms, a topological search for the (3,-3) critical points in the negative Laplacian has been
performed and the resulting VSCC are listed plotted in Figure 3.9 for the theoretical density. Only
the theoretical results are shown here whereas the experimental ones can be found in the Supporting
Information. The results reveal the presence of three maxima pointing away from the S-O bond for
all oxygen atoms. These are usually indicated as non-bonding maxima (NBM) and are associated
with lone pairs in the Lewis theory. An additional (3,-3) CP is found for O2 along the S-O
interaction line, while one (3,-3) maximum is found along each S-O bond, located in the O basin ,
close to the S-O bcp and the S O interatomic surface. These maxima, lying along the bond paths,
are referred to as bonding maxima (BM). The presence of these BM and NBM accounts for the sp3
hybridization of the oxygen atoms, thus in agreement with the picture of a single and, because of
their location, polarized S-O bond. However, as said before, three NBM would also be compatible
with the situation in which some mixing of the two forms in Figure 3.1 occurs and possibly even for
the ensemble of the resonant forms corresponding to the pure hypervalent description. For the S-O1
and S-O3 interactions, only one BM is found between these atoms, probably due to enhanced
electron sharing. This is at first surprising since one would expect such a situation for the shortest
interaction, which is S-O2, but it may be easily rationalized in consideration of the fact that a bond
length decrease leads to a more polarized S-O bond (see infra).
An almost identical picture is obtained from the experimental data. The only major difference is that
for O2 only two VSCC regions pointing away from sulfur are found. This is to be expected since, as
seen from Figure 3.7, they are both positioned in the mirror plane. If an additional maximum was to
be found outside of the plane this would have to appear twice (due to the mirror plane symmetry)
giving a total of 5 VSCCs for O2 (including the BM). This discrepancy between theory and
experiment needs a further analysis, not yet concluded at the time of writing this thesis. It requires
49
to evaluate the value of the Laplacian at the saddle point between the two maxima, and use an
isocontour value just above the saddle value (that is a slightly more negative Laplacian value). This
should, at least pictorially, highlights the presence (or confirm the absence) of the “missing” NBM.
However, even if missing, there is still the possibility that the two maxima have merged in just one
maximum and one irreducible domain, which may be due to a greater weight of the resonant form
with a “double bond” or to an enhanced sphericalization of the O basin.
Figure 3.9 Bonded (BM) and non-bonded (NBM) maxima for the S-O interactions. The small blue spheres represent the maxima found within the oxygen VSCCs and close to the oxygen nucleus (∼0.35Å from the O nucleus) while the small orange spheres are the ones lying along the S-O bond paths, closer to the sulphur (∼0.74Å from the S nucleus) but still located in the oxygen basins. The small black spheres mark the position of the B12, B13 and B14 BCPs.
Table 3.2 lists parameters obtained from atomic basin integrations of all unique atoms both for the
experimental and the theoretical data. As expected the net charges of the potassium atoms are close
to +1. The experimentally obtained Bader net atomic charge on each of the four oxygen atoms is on
average around -1.4 electrons and the net charge on sulfur is around +4.3 electrons giving a value as
large as -1.36 e for the sulfate group. Comparing this with the basin integrations of the density
50
obtained from theoretical calculations the Bader net atomic charges obtained are very similar. The
net charge on S is around +4 which is slightly higher (∼0.4) than the experimental value. Similarly
the potassium net charges obtained from theory are slightly higher at ∼0.9 compared with ∼0.75
from experiment. Cioslowski and Surján interpret their large Bader atomic charges in S-O
containing compounds as proof of a large degree of ionicity in the bonds.44 Our results point in the
same direction with the knowledge that the huge charge separation is actually due to strong
polarization rather than an ionic description of the bond.
Table 3.2 Bader’s electron populations, N(Ω), and net charges, q(Ω) for the experimental and theoretical charge densities .
Atom Ω exp theo
K1 N(Ω) 18.28 18.1 q(Ω) 0.72 0.90
K2 N(Ω) 18.23 18.1 q(Ω) 0.77 0.90
S N(Ω) 11.73 12.14 q(Ω) 4.27 3.86
O1 N(Ω) 9.37 9.42 q(Ω) -1.37 -1.42
O2 N(Ω) 9.39 9.41 q(Ω) -1.39 -1.41
O3 N(Ω) 9.43 9.42 q(Ω) -1.43 -1.42
As a final characterization of the bonding features in K2SO4 using only electron density based
descriptors, the Source Function (SF) contributions from the various atomic basins have been
evaluated for the theoretical electron density.50 In Table 3.3 the percentage contribution, %SF, to
the electron density at the various S-O BCPs is reported.
51
Table 3.3 %SF contributions to S-O BCPs. The symbol O3σ identify the oxygen atom equivalent to O3 due to the presence of the mirror plane.
BCP Bond S O1 O2 O3 O3σ TOT B14 S-O1 28.5 49.4 6.8 6.7 6.7 98.2 B12 S-O2 29.5 6.4 50.0 6.5 6.5 98.9
B13 S-O3 28.6 6.6 6.8 49.6 6.6 98.3
S- O3σ 28.6 6.6 6.8 6.6 49.6 98.3
As can be seen about 79% of the density at the BCP is determined by the sulfur and its linked
oxygen atom. This value is comparable to that found for a prototypical covalent bond such as the
C-C single bond in ethane50b and indicates that the valence charge involved in bonding is localized,
though not as much as found in conventional covalent double bonds (%SF around 85-90%). The
large difference between the %SF values for S and O, with the linked O determining almost twice
the density contributed by the S atom at the bcp, clearly denotes the largely polarized nature of the
S-O shared bond. Not to be neglected is the density contribution determined by the oxygen atoms
within the same sulfate group not directly involved in the bond (Table 3.4). Such cumulative
contribution amounts to about 20% of the S-O bcp density, that is about 2/3 of the density
determined at the same point by the sulphur atom. This is perhaps an unexpected result in view of a
single-bonded depiction of the bonding in the sulfate ion (Figure 3.1, right) and one could naively
tempted to ascribe it just to the proximity of the oxygen atoms and to the inverse dependence from
the distance to the reference point in the local source function expression.50 However, the
importance of the neighboring oxygen atoms is neatly corroborated by the appreciably large
electron exchange existing between oxygen atoms, as expressed by the values of the delocalization
indices50b δ(O,O’). Evaluation of such indexes requires the pair density, not available at the
experimental level, and expressed only in terms of the first order density matrix within a single
determinant approach (at Kohn-Sham DFT level the approximate formula that makes use of an
HF-like second-order exchange density matrix is customarily adopted). Table 3.4 reports %SF and
52
δ(O,O’) values calculated for a Td molecular model of the sulphate group, at the equilibrium S-O
distance (B3LYP/6-311G++(d,p) level) and for an elongated and a shortened S-O bond distance.
Reported in this table are also the net charge on the oxygen atom and the value of the Laplacian at
one of the three equivalent oxygen lone pairs concentrations nbm, as they both change with
distance. Before discussing data in Table 3.5, we just recall that the number of electrons on an atom
Ω, N(Ω), may be seen as given by the sum of those who are localized within the basin, λ(Ω),
∑Ω≠Ω
ΩΩ+Ω=Ω'
),'(21
)()( δλN Eq. 1
and those that are delocalized (exchanged) between them (second term in the right-hand expression
of Eq. 1). For typical single, double and triple covalent bonds, like the CC bond in ethane, ethene,
ethine δ(C,C’) are almost equal to one, two and three. When a bond is largely polarized, electrons
are less effectively exchanged and, for instance, a δ(C,O) value of about 1.6-1.8 is observed for the
C≡O molecule. Let’s so first consider the S-O bond at the equilibrium distance: δ(S,O) is 0.96 and S
and O are exchanging about one electron as expected for a covalent single bond or for a highly
polarized single bond with partial double bond character. The number of electrons delocalized
between each pair of oxygen δ(O,O’) = 0.23 is not negligible and the sum of electrons each oxygen
exchanges with the neighboring oxygen atoms in the sulphate group is only 20-30% less than the
electron charge it exchanges with the sulphur atom to which is directly bonded! When the S-O bond
is forced to shorten or elongate, the contribution of the various resonant forms of the hypervalent
bonding scheme (Fig. 3.1, left) is expected to respectively increase and decrease and a similar trend
should thereby be observed as for the capability of the oxygen atoms to exchange electrons between
themselves.e
e Note that the evaluation of the ellipticity at the bcp, or at any point along the bond path, does not
provide information about double-bond character in such case, since Td symmetry implies a perfect cylindrical symmetry around the bond, regardless of its actual bond order. A similar
53
Table 3.4 %SF contributions for the S-O BCP in a Td model for the sulphate group at equilibrium (1.53Å), shortened and elongated bond distance. Delocalization indeces δ, the net oxygen charge q(O) and the value of the negative of the Laplacian density at one of the three equivalent oxygen lone pairs concentrations, NBM, is also reported for the various S-O distances.
d (Å) ρ (e/Å5)
%SF (S) %SF (O) ΣO’%SF (O’) δ(S,O) ΣO’ δ(O,O’) q(O) NBM (e/Å5)
1.47 1.92 27.1 49.8 21.6 0.87 0.74 -1.46 92.8 1.53 1.78 31.8 48.0 20.1 0.96 0.68 -1.37 96.2 1.59 1.63 36.9 45.3 17.4 1.08 0.61 -1.23 100.3
Indeed (Table 3.4) the bcp density is largest for the shortest distance and smallest for the elongated
geometry, suggesting an enhanced “double-bond” character with S-O distance shortening.
Furthermore, the trend of the δ(O,O’) values indicates that the capability of the oxygen atoms to
exchange electrons between oxygen atoms increases with decreasing S-O distance and an analogous
trend is observed for the %SF contribution from neighboring oxygen atoms to the S-O bcp density.
The increased availability of electrons in the bonding region with decreasing S-O distance is also
revealed by the parallel decrease of the electronic charge concentration of the lone pairs as indicated
by the significant decrease of the magnitude of the -∆∇2ρ NBMs. However, one also note that
unexpectedly the electron exchange between S and O (and thus the bond order) largely diminishes
with decreasing S-O distance, rather than increasing as also corroborated by the decreasing
percentage contribution from the S and O atoms to the density at their intervening bcp. The δ(S,O)
value diminishes from 1.08 at R equal to 1.59 Å to 0.87 at R equal to 1.47 Å, while the sum of %SF
contributions from the oxygen and the sulphur atoms show a parallel decrease from 82.2 to 76.9%.
What is very interesting is to observe that the decrease of δ(S,O) and of the %SF(S+O) contribution
situation holds also for any situation which is a slight perturbation with respect to the Td symmetry, as it is the case of our experimental geometry .
54
is accompanied by a clear increase of the bond polarity, since the ratio of sources from sulphur and
oxygen atoms lowers from 0.81 at R equal to 1.59 Å to only 0.54 at d equal to 1.47 Å. Note that an
homopolar bond would have a ratio of 1 at any distance. In other words, the SF not only mirrors
the changes in the electron density exchange described by the delocalization indices, but also
provides a rationale for such changes. The unexpected electron delocalization decrease with
decreasing distance is the result of a strongly enhanced bond polarization, as also indicated by the
largely increased charge transfer with decreasing distance (Q(O) values in Table 3.4).
Overall, when forcing an increase of the contribution of the various resonant forms of the
hypervalent bonding scheme, the S-O bond order is found to decrease rather than increase because
of the counterbalancing effect due to the largely enhanced bond polarity. The electronic charge
withdrawn from the lone pair regions and made available in the S-O bonding region turns out yet to
be essentially localized on the oxygen basins, a fact which explains why oxygen atoms increase
their ability to exchange electron among themselves, despite their decreased electron exchange with
the sulphur atom. This mechanism while clearly ruling out any hypervalent description of the sulfur
atom in the sulfate group, does not deny the possibility of an appreciable electron exchange among
the oxygen atoms of this group, due to a partial contribution of resonant forms having a so polarized
“double-bond” character that the actual S-O bond order at equilibrium does not exceed that of a
standard covalent single bond.
4. Conclusions
The S-O bonding has been investigated through a study of the charge density of the sulfate group,
SO42-, within a crystalline environment, using both experimental and theoretical electron densities.
The experimental charge density was determined by multipole least squares refinement, while a
theoretical density was obtained from periodic ab-initio DFT calculations. The chemical bonding
55
was analysed within the framework of the Quantum Theory of Atoms In Molecules (QTAIM). The
Source Function topological descriptor has also been applied to provide further insight on the
electron distribution within the K2SO4 crystal lattice. The combined evidence suggests a bonding
situation where the S-O interactions can be characterized as highly polarized, covalent bonds, with
“single bond” description significantly prevailing over the “double bond” picture. Model
calculations on systems with shortened S-O bonds with respect to equilibrium, indicate that even
when the lone pair electrons on oxygen atoms are forced to participate to bonding (formally
enhancing the “double bond” character), an unexpected decrease in the S-O electron sharing is
observed since a parallel increase in the charge transfer between the sulphur and the oxygen atoms
occurs, leading to a further polarization of the S-O interactions. Conversely, the non negligible
electron sharing among oxygen atoms already present at equilibrium, is further enhanced upon S-O
bond shortening. Overall, this picture rules out the possibility of a hypervalent description of the
sulfur atom in the sulfate group, despite the appreciable electron exchange occurring among the
oxygen atoms of this group.
56
Chapter 4
Coordination Polymers: the case of [Zn(HCOO)2(H2O)2]∞∞∞∞
1. Introduction
Coordination polymers (CPs) have received an enormous attention due to their broad range of
properties ranging from gas-storage, porous properties, catalytic properties, negative thermal
expansion and magnetic properties. 51 Unlike many other porous materials the reactants are
preserved during the synthesis and thus, it is in principle possible to tailor the product by carefully
choosing the reactants. However, it has been shown that minor changes in synthesis parameters can
lead to different products,52 and often the synthesis can lead to multiple products.53
A lot of experimental work has been already carried out at Århus university on this class of
compounds. The studies have been mainly concerned with the magnetic properties of coordination
polymers through analysis of the X-ray charge densities.54 These analyses are often complicated by
poor crystal quality and quite complicated structures which often include disordered solvent
molecules.55
The isostructural series have been known since the 1960s56 and have been studied mainly due to
their interesting magnetism, which, despite a wealth of studies, is not completely understood and
inconsistencies still exists. The most studied compound in the series is the Mn containing analog
which shows at least three phase transitions at 3.7, 1.7 and 0.6 K.57 The various experiments
revealed that the Mn(1) layers order in a canted antiferromagnetic sublattice at 3.7 K. At 1.7 K this
sublattice spontaneously reorients. The Mn(2) sublattice follows a Brillouin like function with only
slowly increasing magnetization below the Néel temperature,58 and it has therefore often been
referred to as being paramagnetic down to 0.6 K. It has been suggested that the formate linkers act
as superexchange pathways for the magnetic interaction.
57
The work reported here couples with the experimental analysis performed by Mads R. V. Jørgensen
(PhD stud.) and is focused on one of the simplest coordination polymers known,
Zn(HCOO)2(H2O)2. Due to the presence of Zn, the compound is nonmagnetic and can thus be
considered a reference structure aiding the understanding of the magnetic compounds in the series.
2. Crystallographic structure
The structure of the polymer is shown in Figure 4.1. The compound crystallizes in the monoclinic
space group P21/c and is composed of Zn, formic acid residues, and water molecules. There are two
distinct metal sites in the structure both located on sites with inversion symmetry. Both are
coordinated in a slightly distorted octahedral field. Zn(1) is coordinated to six formate ligands, four
of these create layers in the bc-plane, and all Zn atoms in these layers are interconnected with
formate linkers. Zn(2) is coordinated to two formate ligands and four water molecules. This creates
layers of mutually unconnected Zn(2) ions interpenetrating the layers formed by Zn(1) (Figure 4.2).
The two kinds of layers are connected by formate linkers from Zn(1) to Zn(2) creating a 3D
network. The hydrogen atoms on the two water molecules are all involved in hydrogen bonding
(Figure 4.3). Three of these provide interaction between the Zn(1) and Zn(2) layer, while the last,
shown in red in the figure, provides a contact within the Zn(2) layer.
Figure 4.1 Structure of Zn(HCOO)2(H20)2 at 100K. The labels are only shown for atoms in the asymmetric unit. Thermal ellipsoids at 50% probability level.
58
Figure 4.2 Projection along the b-axis of the 2x2 super cell.
Figure 4.3 Section of the polymer structure. The H-bonds within the Zn(2) layer are shown in red.
3. Computational details
The wave function was calculated using ab initio periodic density functional theory employing
B3LYP exchange and correlation functional (Becke, 1993; Lee et al., 1988) implemented in the
program CRYSTAL06.16 The basis set used in the calculations was a standard Pople 6-311G basis
set for all atoms except Zn. The basis set for Zn was a def2-TZVP basis set,18 overall corresponding
to an all electron triple-ζ basis. To avoid issues related to quasi-linear dependence the coefficients
59
of the outermost s and p functions has been recalculated. The f functions have furthermore been
removed. Two different geometries have been determined experimentally, based on data collected
at 15K and at 100K. Both the geometries have been considered for the theoretical modelling. The
energy was minimized and a criterion of ∆E < 10-6 Hartree was used to ensure that the model has
reached convergence. The topology of the density was analyzed using the beta version of the
program TOPOND08.48
4. Results and Discussion
As for K2SO4, no results based on the experimental work performed by Mads R. V. Jørgensen (PhD
stud.) will be reported here, unless fundamental for the ongoing discussion. In any case mainly a
qualitative view will be given.
To accurately model the Zn atoms from the collected data, several models were tested and the
residual density was scrutinized for systematic features. In previous charge density studies of Zn
containing compounds the analysis were complicated by anharmonic motion, which can correlate
with the multipole parameters, thus leading to difficulties separating vibration and bonding features.
This could lead to a masking of the detailed bonding features around the presumably highly
spherical Zn atom. However, in this study there were no indications of anharmonic motion of the Zn
atom.
From overall charge considerations the charge of each Zn atom in Zn(HCOO)2(H2O)2 is expected
to be in a formal +2 valence state. Therefore, a plausible model would be to keep the 3d10 shell
unperturbed and refine population and possible aspherical features of the 4s electrons. Alternatively
the 4s electrons could be removed completely from Zn and distributed to the surrounding oxygen
atoms. This could be combined with the use of an ionic radial function. In the isostructural
compound Mn(HCOO)2(H2O)2 it was found that there were some electron donation from the
60
ligands into the 3d orbitals thus leading to some perturbation of the atomic orbitals.59 By analogy
the ligands could potentially perturb the valence shell, thus lead to an aspherical charge distribution,
not only of the 4s electrons, but also of the 3d electrons. Therefore it would be important to test if
the 3d electrons were indeed perturbed. Five different models denoted I-V, with neutral radial
functions from the SCM data bank and including 4s electrons were tested against the 100 K data.
Some results are shown in Table 4.1.
Table 4.1 Comparison of different Zn charge density models. I: 3d population and multipoles refined, 4s population fixed to 2e- and spherical. II: 3d population fixed at 10e- and 4s population refined spherically. III: as II but 4s refined aspherically. IV: 3d and 4s monopole populations refined, but only 3d multipoles refined. V: As IV but in addition 4s multipoles were refined. Pxy is the population of the xy orbital, R(F2) is the R-value, ∆ρ is peak value in the residual density and QΩ is the integrated charge of the atomic basin.
The 4s electrons of transition metals are notoriously hard to model due to their diffuse nature, and
thus often only a few low order reflections will contain information about these electrons. These
low order reflections are often weakened by extinction, which have been refined to a significant
value for both data sets. The differences between the models are minor, yet a careful analysis of the
results seemed to indicate model I as the one best describing the data. This was quite surprising,
since it includes a filled 4s orbital in contrast to the expected +2 valence state for Zn. Also
surprising was the fact that deformation densities still revealed that the Zn atoms were not
completely spherical. It was encouraging to find that also the deformation densities maps obtained
61
from the theoretically derived electron density showed a similar structure around the metal atom
(Figure 4.4).
Figure 4.4 Deformation densities maps for the experimental (left) and theoretical (right) models. Contours are plotted at 0.1eÅ-3, the zero contour is omitted. Blue regions represent positive values of ∆ρ, while red regions negative values.
Of course a direct comparison of the deformation densities from the experimental models and the
theoretical model is not possible as the reference models, superposition of neutral spherical atoms,
are not identical and thus, only a qualitative comparison can be made. Still it is interesting to note
that, besides acknowledging the asphericity of the Zn atoms, both theory and experiment highlights
accumulation of density directed towards the negative ligands in the presumably rather ionic Zn-O
bonds, even more so as the largest accumulations are found along the shorter ones.
Indeed this could be a sign of covalent character, thus supporting the idea that shared shell
interaction are present in this compounds. However, from the topological analysis of the electron
density this view seems to be lacking. For both the Zn atoms three different interactions with the
surrounding oxygen atoms have been found (Table 4.2).
62
Table 4.2 Topological interactions for the Zn-O bonding interactions.
Bond d (Å) AILA-bcp (Å) AILbcp-B (Å) ρ(e/Å3) ∇2ρ (e/Å5) ε
Zn1-O4 2.0608 1.0017 1.0604 0.440 8.446 0.037 Zn1-O3 2.0957 1.0185 1.0789 0.402 7.611 0.032 Zn1-O1 2.1302 1.0341 1.0980 0.372 6.793 0.029 Zn2-O5 2.0546 0.9985 1.0564 0.438 8.729 0.069 Zn2-O6 2.0970 1.0154 1.0819 0.409 7.700 0.055 Zn2-O2 2.1644 1.0518 1.1153 0.336 6.124 0.017
In all the cases a decrease of the value of ρ is observed according to an increase of the bonding
distance. The Laplacian of ρ, ∇2ρ, is found to be positive for every of such interactions, suggesting
that the nature of the M-O interaction is closed-shell. However, since the N shell is not revealed by
the Laplacian for the considered metals,60 this value can’t be considered a definitive proof. Further
insights can be obtained plotting the ∇2ρ profiles along the line joining the bonded atoms and
comparing the value of the Laplacian, at the BCP, with the one that is obtained from the IAM
(Independent Atoms Model) description.
This is shown in Figure 4.5 for the Zn2-O5 interaction (similar plots are found for all the Zn-O
bonds) and it is evident that, upon bond formation, charge density is removed from the bonding
region resulting in the positive value found for the Laplacian. Figure 4.5 also rules out the
hypothesis of a highly polarized covalent bond, such as found for the S-O interactions (see Chapter
3)
A different picture arises for the C-O, C-H and O-H interactions, for which the expected shared-
shell nature of the bond is found. This is supported by the high value of ρ and the negative value of
∇2ρ at the BCP (Table 4.3 and Table 4.4). Furthermore, the Laplacian profiles (Figure 4.6 and
Figure 4.7) shows charge accumulation between the two atoms involved in the bond and, for C-O
and O-H, the shift of the BCP towards the most electronegative atom reveals some degree of
polarization.
Figure 4.5 VSCCs around Zn(1) (left) and Zn(2) (right). Magenta spheres showthe experimental density at 100 K and blue spheres the positions from the theoretical density. Bonds correspond to the interatomic ve
Table 4.3 Topological parameters for the ligands.
Bond d (Å) AILA
C1-O1 1.2646 0.4554C1-O2 1.2555 0.4515C1-H1 1.0840 0.7186C2-O3 1.2530 0.4498C2-O4 1.2605 0.4537C2-H2 1.0840 0.7159
Table 4.4 Topological parameters for the
Bond d (Å) AILA
O5-H3 0.9900 0.7715O5-H4 0.9900 0.7671O6-H5 0.9900 0.7671O6-H6 0.9900 0.7667
VSCCs around Zn(1) (left) and Zn(2) (right). Magenta spheres showthe experimental density at 100 K and blue spheres the positions from the theoretical
Bonds correspond to the interatomic vectors.
Topological parameters for the C-O and C-H bonding interactions in the formate
A-bcp (Å) AILbcp-B (Å) ρ (e/Å3) ∇2ρ
0.4554 0.8093 2.331 -13.0250.4515 0.8042 2.383 -12.8690.7186 0.3530 1.871 -20.6390.4498 0.8032 2.397 -12.7170.4537 0.8069 2.360 -13.1810.7159 0.3560 1.872 -20.511
Topological parameters for the O-H bonding interactions in the water ligands.
A-bcp (Å) AILbcp-B (Å) ρ (e/Å3) ∇2 ρ
0.7715 0.2005 2.120 -44.0580.7671 0.2051 2.136 -43.3710.7671 0.2052 2.143 -43.5540.7667 0.2057 2.147 -43.296
63
VSCCs around Zn(1) (left) and Zn(2) (right). Magenta spheres show positions from the experimental density at 100 K and blue spheres the positions from the theoretical
bonding interactions in the formate
ρ(e/Å5) ε
13.025 0.058 12.869 0.044 20.639 0.009 12.717 0.046 13.181 0.052 20.511 0.008
eractions in the water ligands.
ρ (e/Å5) ε
44.058 0.019 43.371 0.019 43.554 0.019 43.296 0.019
Figure 4.6 Laplacian profile for the displayed on the left, while O/H atom is on the right. The dashed vertical line marks the position of the BCP
Figure 4.7 Laplacian profile for the O5displayed on the left, while H atom is on the right. The dashed vertical line marks the position of the BCP. The inset shows an enlargement of the region next to the BCP
The last kind of interactions revealed by the topological analysis of the electron density
hydrogen bond. They can be divided in two categories: O
oxygen and water hydrogen, and C
oxygen (Table 4.5 and Table 4.6). As illustrated in the introduction of this chapter, the first ones
(the shortest) are the most interesting ones to analyze due to the possibility of their involvement in
the propagation of the magnetic information in the structure.
Laplacian profile for the C1-O1 (left) and C1-H1 (right) interactions. The C atom is displayed on the left, while O/H atom is on the right. The dashed vertical line marks the position of the BCP
Laplacian profile for the O5-H3 (left) and (O2-H3) interactions. The O atdisplayed on the left, while H atom is on the right. The dashed vertical line marks the position of the BCP. The inset shows an enlargement of the region next to the BCP
The last kind of interactions revealed by the topological analysis of the electron density
. They can be divided in two categories: O-H⋅⋅⋅O taking place between formate ions
oxygen and water hydrogen, and C-H⋅⋅⋅O taking place between formate ions hydrogen and water
oxygen (Table 4.5 and Table 4.6). As illustrated in the introduction of this chapter, the first ones
(the shortest) are the most interesting ones to analyze due to the possibility of their involvement in
n of the magnetic information in the structure.
64
H1 (right) interactions. The C atom is displayed on the left, while O/H atom is on the right. The dashed vertical line marks
H3) interactions. The O atom is displayed on the left, while H atom is on the right. The dashed vertical line marks the position of the BCP. The inset shows an enlargement of the region next to the BCP.
The last kind of interactions revealed by the topological analysis of the electron density is the H⋅⋅⋅O
O taking place between formate ions
tween formate ions hydrogen and water
oxygen (Table 4.5 and Table 4.6). As illustrated in the introduction of this chapter, the first ones
(the shortest) are the most interesting ones to analyze due to the possibility of their involvement in
65
Table 4.5 Topological parameters for the O-H⋅⋅⋅O interactions BCPs
Bond d (Å) AILA-bcp (Å) AILbcp-B (Å) ρ(e/Å3) ∇2 ρ (e/Å5) ε
O2⋅⋅⋅H3 1.7315 1.1428 0.6074 0.291 3.467 0.033 O1⋅⋅⋅H6 1.7649 1.1564 0.6257 0.283 3.316 0.025 O4⋅⋅⋅H5 1.7723 1.1597 0.6320 0.264 3.292 0.034 O3⋅⋅⋅H4 1.7807 1.1647 0.6344 0.261 3.234 0.038
Table 4.6 Topological parameters for the various C-H⋅⋅⋅O interactions BCPs.
Bond d (Å) AILA-bcp (Å) AILbcp-B (Å) ρ(e/Å3) ∇2 ρ (e/Å5) ε
O3⋅⋅⋅H2 2.3580 1.3884 1.0162 0.090 1.260 0.464 O1⋅⋅⋅H2 2.4859 1.4716 1.0405 0.072 0.901 0.215 O4⋅⋅⋅H1 2.5620 1.5035 1.0986 0.058 0.781 0.458 O5⋅⋅⋅H1 2.7248 1.6481 1.0899 0.040 0.488 0.021 O6⋅⋅⋅H1 2.9005 1.6866 1.2801 0.031 0.430 0.241
If the analysis of the bonding interactions is conducted considering also the electron energy
density 61 it can be seen how, rather than simply closed shell, all the Zn-O and the H-bond
interactions belong to the so-called “transit region” where ionic and covalent character are
somewhat competitors.
Prompted by this new result, the Source Function has been evaluated to gain further insight on this
kind of interaction. While it was hidden in the “simple” analysis based on the sign of the Laplacian,
through the SF it is found that, for every Zn-O interaction, the atoms directly involved in the
bonding provides only 60% of the total electron density which is recovered at the BCP (in almost
equal amounts, ~30%). The remaining is actually provided by all the other atoms, including carbons
contributing for ~10%. While this analysis is still ongoing, it can be a definite sign of the existence
of a π-delocalization scheme connecting the Zn atoms.
5. Conclusions
The charge density of the coordination polymer Zn(HCOO)2(H2O)2 has been modeled based on two
X-ray diffraction data sets obtained at 15 and 100 K. Furthermore, densities from periodic DFT
66
calculations based on the experimental geometry have been obtained. In general there is a good
agreement between the two experimental and the theoretical densities. The charge density of Zn in
various compounds reported in the literature and the description is quite varying from highly
spherical to distorted ions. The charge density of the title compound show overall quite spherical Zn
atoms. However differences between the two experimental descriptions are found. It is clear that the
description of Zn features require extremely high quality data as several models yields similar
residuals but varying description of the charge density. The deformation density maps show
accumulations towards the oxygen atoms of the ligands. This is also found in the theoretical models
and supports the idea of covalent character in this interactions. This is also strengthened by the
analysis of the electron energy density and of the Source Function.
67
Chapter 5
Conclusions and Future Work
The charge density of FeX2 marcasitic compounds, of crystalline K2SO4 and of a reference
compound for magnetically active coordination polymers have been analysed from both the
experimental and crystallographic point of view. In the two cases where the work took the moves
from the need to disentangle the ambiguities arising from the experimentally derived electron
density, ab-initio calculations proved to be able to aid, at least qualitatively. In many cases,
especially through the topological analysis of ρ, the role of theory has been also a quantitative one.
This was, for example, the case of K2SO4. The presence of heavy elements and extinction effects
masked the fine mechanisms related to the charge polarization within the oxygen atoms themselves.
Of course also the opposite is true, as that of theory and experiment in charge density analysis is a
synergic and strongly intertwined work. A finer analysis of the multipole-modeled ρ would
hopefully provide support to the mechanism governing the bonding interactions in FeX2
compounds.
Even if almost completed, the analysis of Zn(HCOO)2(H2O)2 just scratched the surface of what’s
hidden within this family of compounds. Presently, analyses on the Mn and Fe containing
compounds are in an advanced state.
Lastly, even if unfortunately they haven’t found space in these pages, two others projects should be
mentioned, both dealing with crystals characterized by high disorder. The first one, in collaboration
with PhD student Mette Søberg and postdoc Henrik F. Clausen (Århus university), is aimed at
unraveling the properties of hydroquinone giving rise, when in its crystal form, to a host-guest
structure able to imprison small solvent molecules like methanol and acetonitrile. The second one,
68
in collaboration with Dr. Marlena Gryl (Jagiellonian university), is aimed at evaluating the bonding
features in co-crystals of organic molecules characterized by non-linear optical properties.
69
References
[1] Bentien, A.; Johnsen, S.; Madsen, G. K. H.; Iversen, B. B.; Steglich, F. Eur. Phys. Lett. 2007,
80, 17008-17012
[2] Bentien, A.; Madsen, G. K. H.; Johnsen, S.; Iversen, B. B., Phys. Rev. B. 2006, 74, 205105
[3] Sun, P.; Søndergaard, M.; Sun, Y.; Johnsen, S.; Iversen, B. B.; Steglich, F. Appl. Phys. Lett.
2011, 98, 072105; Sun, P.; Søndergaard, M.; Iversen, B. B.; Steglich, F. Ann. Phys. 2011,
523, 612–620; Sun, P.; Oeschler, N.; Johnsen, S.; Iversen, B.B.; Steglich, F., Dalton Trans.
2010, 39, 1012-1019
[4] Holseth, H.; Kjekshus, A. Acta Chemica Scandinavica 1968, 22, 3284-3292
[5] Dahl, E. Acta Chemica Scandinavica 1969, 23, 2677-2684
[6] Brostigen, G.; Kjekshus, A. Acta Chemica Scandinavica 1970, 24, 1925-1940
[7] Brostigen, G.; Kjekshus, A. Acta Chemica Scandinavica 1970, 24, 2983-2992
[8] Brostigen, G.; Kjekshus, A.; Römming, C. Acta Chemica Scandinavica 1973, 27, 2791-2796
[9] xxx Acta Chemica Scandinavica. A28 (1974) 996
[10] xxx Acta Chemica Scandinavica. 24 (1970) 2993
[11] xxx Acta Chemica Scandinavica. 24 (1970) 3309
[12] Goodenough, J.B. Journal of Solid State Chemistry 1972, 5, 144-152
[13] Hullinger, F.; Mooser, E. Journal of Physics and Chemistry of Solids 1965, 26, 429-433
[14] Angew. Chem. Int. Ed. 39 (2000) 2408
[15] J. Mater. Chem. 18 (2008) 2090
70
[16] Dovesi, R.; Saunders, V.R.; Roetti, C.; Orlando, R.; Zicovich-Wilson, C.M.; Pascale, F.;
Civalleri, B.; Doll, K.; Harrison, N.M.; Bush, I.J.; D’Arco, Ph.; Llunell, M. CRYSTAL06
User’s Manual, University of Torino, Torino, 2006
[17] Weigend, F.; Ahlrichs, R. Physical Chemistry Chemical Physics 2005, 7, 3297
[18] Ahlrichs, R.; May, K. Physical Chemistry Chemical Physics 2000, 2, 943
[19] M. Dolg, U. Wedig, H. Stoll, H. Preuss, J. Chem. Phys. 86, 866 (1987); J.M.L. Martin, A.
Sundermann, J. Chem. Phys. 114, 3408 (2001); B. Metz, H. Stoll, M. Dolg, J. Chem. Phys.
113, 2563 (2000)
[20] Perdew, J. P.; Burke, K.; Ernzerhof, M. Physical Review Letters 1996, 77, 3865
[21] a) Becke, A. D. Journal of Chemical Physics 1993, 98, 5648; b) Perdew, J. P. Electronic
Structure of Solids, Akademie Verlag, Berlin (1991); c) Perdew, J. P.; Wang, Y. Physical
Review B 1986, 33, 8800; d) Perdew, J. P.; Wang, Y. Physical Review B 1989, 40, 3399; e)
Perdew, J. P.; Wang, Y. Physical Review B 1992, 45, 13244
[22]Boda, G.; Stenstrom, B.; Sagredo, V.; Beckman, O. Physica Scripta 1971, 4, 132-134
[23] Sithole, H.M.; Nguyen-Manh, D.; Pettifor, D.G.; Ngoepe, P.E. Molecular Simulation 1999,
22 31-37
[24] App. Phys. Express 2 (2009) 091102
[25] Eur. Phys. J. B 53 (2006) 205
[26] Phys. Rev. B 74 (2006) 205105
[27] App. Phys. Express 2 (2009) 091102
[28] Phys. Rev. B 74 (2006) 205105
[29] Bader, R. F. W. Atoms in molecules. A quantum theory. 1990, Oxford University Press,
Oxford, UK
71
[30] Bianchi, R.; Gatti, C.; Adovasio, V.; Nardelli, M. Acta Cryst. B 1996, 52, 471–478.
[31] Macchi, P.; Sironi, A. Coordination Chemistry Review 2003, 238, 383-412
[32] Gatti, C. Z. Krist. 2005, 220, 399-457
[33] Cortéz-Guzmán, F. ; Bader, R.F.W. Coordination Chemistry Review 2005, 249, 633-662
[34] Scherer, W.; McGrady, G. S. Angew. Chem. Int. Ed. 2004, 43, 1782-1806
[35] Saito, Y. Inorganic Molecular Dissymmetry. Springer-Verlag, 1979 New York
[36] Koritsanszky, T. S.; Coppens, P. Chem. Rev. 2001, 101, 1583–1627.
[37] Bader, R.F.W. Coord. Chem. Rev. 2000, 197, 71–94.
[38] Gillespie, R.J. Canadian Journal of Chemistry 1992, 70, 742-750
[39] Lewis, G.N. J. Am. Chem. Soc. 1916, 38, 762-785.
[40] (a) Cioslowski, J.; Mixon, S.T. Inorg. Chem. 1993, 32, 3209-3216. (b) Wang, Y.; Chang, N.
L.; Pai, C. T. Inorg. Chem. 1990, 29, 3256-3259. (c) Lee, C. R.; Tang, T.-H.; Chen, L.; Wang,
Y. Chem. Eur. J. 2003, 9, p. 3112-3121. (d) Bats, J.W.; Coppens, P.; Koetzle, T. F. Acta
Cryst. 1977, B33, 37-45. (e) Figgis, B. N.; Khor, L.; Kucharski, E. S.; Reynolds, P. A. Acta
Cryst. 1992, B48, 144-151. (f) Figgis, B. N.; Iversen, B. B.; Larsen, F. K.; Reynolds, P. A.
Acta Cryst. 1993, B49, 794-806.
[41] (a) Dobado, J. A.; Martínez-García, H.; Molina, J.; Sundberg, M. R.; J. Am. Chem. Soc. 1999,
121, 3156-3164. (b) Dobado, J. A.; Martínez-García, H.; Molina, J.; Sundberg, M. R.; J. Am.
Chem. Soc. 1998, 120, 8461-8471.
[42] Love, I. J. Phys. Chem. A 2009, 113, 2640-2646.
72
[43] (a) Mezey, P. G.; Haas E. C. J. Chem. Phys. 1982, 77, 870-876. (b) Mayer, I. J. Mol. Struct.
THEOCHEM 1987, 149, 81-89. (c) Yu, W.; Guo, L. W.; Lin, H. C.; Kao, C. T.; Tsai, C. J.;
Bats, J. W. Inorg. Chem. 1988, 27, 520-523.
[44] Cioslowski, J.; Surján P. R. J. Mol. Struct. THEOCHEM 1992, 87, 9-33.
[45] Ojima, K.; Nishihata, Y.; and Sawada, A. Acta Cryst. 1995, B51, 287-293.
[46] (a) Clausen, H. F.; Overgaard, J.; Chen, Y. S.; Iversen, B. B., J. Am. Chem. Soc. 2008, 64,
7988-7996. (b) Poulsen, R. D.; Bentien, A.; Chevalier, M.; Iversen, B.B., J. Am. Chem. Soc.
2005, 127, 9156-9166. (c) Coppens, P.; Larsen, F. K.; Iversen, B. B. Coord. Chem. Rev. 2004,
249, 179-195.
[47] Volkov, A.; Macchi, P.; Farrugia, L. J.; Gatti, C.; Mallinson, P.; Richter, T.; Koritsanszky, T.
XD2006 - A Computer Program Package for Multipole Refinement, Topological Analysis of
Charge Densities and Evaluation of Intermolecular Energies from Experimental and
Theoretical Structure Factors; 2006.
[48] Gatti C. TOPOND08 – An electron density topological program for systems periodic in N (N
= 1-3) dimensions.
[49] Morse, M.; Cairns S. S. point theory in global analysis and differential topology; an
introduction. Pure and applied mathematics a series of monographs and textbooks. Academic
Press: New York, 1969.
[50] (a) Bader, R.F.W., Gatti, C. Chem. Phys. Lett. 1998, 287, 233 (b) Gatti, C., Lasi, D. Faraday
Discuss. 2007, 135, 55; (c) Farrugia, L. J.; Macchi, P. J. Phys. Chem A. 2009, 113, 10058-
10067; (d) Gatti, C. Struct. Bond. 2011, 1 $DOI: 10.1007/430_2010_31 2.. (e) Lo Presti, L.;
Gatti C. Chem. Phys. Lett. 2009, 476, 308-316.
[51] Weng, D.-F., Wang, Z.-M. & Gao, S. (2011). Chem. Soc. Rev. 40, 3157-3181; Sculley, J.,
Yuan, D. & Zhou, H.-C. (2011). Energ Environ Sci 4, 2721-2735; Ranocchiari, M. & van
73
Bokhoven, J. A. (2011). Phys. Chem. Chem. Phys. 13, 6388-6396; Meek, S. T., Greathouse, J.
A. & Allendorf, M. D. (2011). Adv. Mater. (Weinheim, Ger.) 23, 249-267; Martinez, C. &
Corma, A. (2011). Coord. Chem. Rev. 255, 1558-1580 Dechambenoit, P. & Long, J. R.
(2011). Chem. Soc. Rev. 40, 3249-3265; Coudert, F.-X., Boutin, A., Jeffroy, M., Mellot-
Draznieks, C. & Fuchs, A. H. (2011). ChemPhysChem 12, 247-258.; Zhou, W. (2010). Chem.
Rec. 10, 200-204; Ma, S. & Zhou, H.-C. (2010). Chem. Commun. (Cambridge, U. K.) 46, 44-
53; Lock, N., Wu, Y., Christensen, M., Cameron, L. J., Peterson, V. K., Bridgeman, A. J.,
Kepert, C. J. & Iversen, B. B. (2010). J. Phys. Chem. C 114, 16181-16186; Peterson, V. K.,
Kearley, G. J., Wu, Y., Ramirez-Cuesta, A. J., Kemner, E. & Kepert, C. J. (2010).Angew
Chem Int Edit 49, 585-588.
[52] Edgar, M., Mitchell, R., Slawin, A. M. Z., Lightfoot, P. & Wright, P. A. (2001). Chemistry –
A European Journal 7, 5168-5175
[53] Clausen, H. F., Poulsen, R. D., Bond, A. D., Chevallier, M. A. S. & Iversen, B. B. (2005). J.
Solid State Chem. 178, 3342-3351.
[54] Poulsen, R. D., Bentien, A., Chevalier, M. & Iversen, B. B. (2005). J. Am. Chem. Soc. 127,
9156-9166.; Poulsen, R. D., Bentien, A., Graber, T. & Iversen, B. B. (2004). Acta Crystallogr
A 60, 382-389; Clausen, H. F., Overgaard, J., Chen, Y. S. & Lversen, B. B. (2008). J. Am.
Chem. Soc. 130, 7988-7996.
[55] Jørgensen, M. R. V., Clausen, H. F., Christensen, M., Poulsen, R. D., Overgaard, J. & Iversen,
B. B. (2011). Eur. J. Inorg. Chem. 549-555; Clausen, H. F., Overgaard, J., Chen, Y. S. &
Lversen, B. B. (2008). J. Am. Chem. Soc. 130, 7988-7996.
[56] Osaki, K., Nakai, Y. & Watanabe, T. (1963). J. Phys. Soc. Jpn. 18, 919.
[57] Flippen, R. B. & Friedberg, S. A. (1963). J. Chem. Phys. 38, 2652-2657; Abe, H. & Matsuura,
M. (1964). J. Phys. Soc. Jpn. 19, 1867-1880; Abe, H., Matsuura, M., Morigaki, H., Yamagata,
74
K. & Torii, K. (1964). J. Phys. Soc. Jpn. 19, 775-776. , R. D. & Friedberg, S. A. (1968). Phys.
Rev. 165, 680-686.
[58] Pierce, R. D. & Friedberg, S. A. (1968). Phys. Rev. 165, 680-686
[59] Poulsen, R. D., Jorgensen, M. R. V., Overgaard, J., Larsen, F. K., Morgenroth, W. G., Graber,
T., Chen, Y. S. & Iversen, B. B. (2007). Chem.--Eur. J. 13, 9775-9790
[60] Shi, Z.; Boyd, R. J.; J. Chem. Phys. 88, 4375 (1988).
[61] Espinosa, E.; Alkorta I.; Elguero, J.; Molins, E.; J. Chem. Phys. 117, 5529 (2002).